Has Protein Folding Been Solved?

Has Protein Folding Been Solved? Sabine Hossenfelder 1.42M subscribers Join Subscribe 18K Share Download Thanks Clip 358,272 views Jan 30, 2021 #biochemistry #education #science NordVPN special deal: Just go to https://nordvpn.com/sabine and use our coupon code sabine at checkout. Recently Deepmind made big headlines with its AlphaFold success. Did it really "solve" protein folding? What did actually happen? In this video I explain what the protein folding problem is, why it's important, and what the current situation is. The protein animation shown at 1 mins 13 seconds goes back to this publication: Structure of the Cdc48 segregase in the act of unfolding an authentic substrate. Cooney I, Han H, Stewart MG, Carson RH, Hansen DT, Iwasa JH, Price JC, Hill CP, Shen PS. Science, 365(6452): 502-505 (2019). https://doi.org/10.1126/science.aax0486 Reused with permission. If you like this video and want more like this, you are helping us a lot if you support us on Patreon: / sabine #science #education #biochemistry 0:00 Intro 0:35 What's the problem? 1:49 Why does it matter? 5:07 How to try and solve it 6:27 The CASP Competition 8:10 Alphafold 2 9:07 What does this mean? 10:42 Sponsor Message Chapters View all Transcript Follow along using the transcript. Show transcript transcript: Intro 0:00 Protein folding is one of the biggest, if not THE biggest problem, in biochemistry. It’s 0:06 become the holy grail of drug development. Some of you may even have folded proteins yourself, 0:12 at least virtually, with the crowd-science app ‘’Foldit”. But then late last year the headlines 0:17 proclaimed that Protein Folding was “solved” by artificial intelligence. Was it really solved? 0:23 And if it was solved, what does that mean? And, erm, 0:28 what was the protein folding problem again? That’s what we will talk about today. 0:32 Has Protein Folding Been Solved? What's the problem? 0:36 Proteins are one of the major building blocks of living tissue, for example muscles, 0:41 which is why you may be familiar with “proteins” as one of the most important nutrients in meat. 0:47 But proteins come in a bewildering number of variants and functions. They are everywhere in 0:53 biology, and are super-important: Proteins can be antibodies that fight against infections, 1:00 proteins allow organs to communicate between each other, and proteins can repair damaged tissue. 1:07 Some proteins can perform amazingly complex functions. For example, pumping molecules in 1:14 and out of cells, or carrying substances along using motions that look much like walking. 1:21 But what’s a protein to begin with? Proteins are basically really big molecules. 1:27 Somewhat more specifically, proteins are chains of smaller molecules called amino acids. But long 1:32 and loose chains of amino acids are unstable, so proteins fold and curl until they reach a stable, 1:39 three-dimensional, shape. What is a protein’s stable shape, 1:43 or stable shapes, if there are several? This is the “protein folding problem”. 1:48 Understanding how proteins fold is important because the function of a protein depends Why does it matter? 1:54 on its shape. Some mutations can lead to a change in the amino acid sequence of a protein 2:01 which causes the protein to fold the wrong way. 2:04 It can then no longer fulfil its function and the result can be severe illness. 2:09 There are many diseases which are caused by improperly folded proteins, for example, type 2:15 two diabetes, Alzheimer’s, Parkinson’s, and also ALS, that’s the disease that Stephen Hawking had. 2:22 So, understanding how proteins fold is essential to figuring out how these diseases come about, 2:29 and how to maybe cure them. But the benefit of understanding protein folding goes beyond that. 2:36 If we knew how proteins fold, it would generally be much 2:39 easier to synthetically produce proteins with a desired function. 2:44 But protein folding is a hideously difficult problem. 2:47 What makes it so difficult is that there’s a huge number of ways proteins can fold. 2:53 The amino acid chains are long and they can fold in many different directions, 2:58 so the possibilities increase exponentially with the length of the chain. 3:04 Cyrus Levinthal estimated in the nineteen-sixties that a typical protein could fold in more than 3:11 ten to the one-hundred-forty ways. Don’t take this number too seriously though. The number 3:17 of possible foldings actually depends on the size of the protein. Small proteins may have 3:23 as “few” as ten to the fifty, while some large ones can have and a mind-blowing ten to the 3:30 three-hundred possible foldings. That’s almost as many vacua as there are in string theory! 3:35 So, just trying out all possible foldings is clearly not feasible. 3:40 We’d never figure out which one is the most stable one. 3:44 The problem is so difficult, you may think it’s unsolvable. But not all is bad. Scientists found 3:51 out in the nineteen-fifties that when proteins fold under controlled conditions, for example in 3:57 a test tube, then the shape into which they fold is pretty much determined by the sequence of amino 4:03 acids. And even in a natural environment, rather than a test tube, this is usually still the case. 4:10 Indeed, the Nobel Prize for Chemistry was awarded for this in 1972. 4:16 Before that, one could have been worried that proteins have a large numbers of stable shapes, 4:22 but that doesn’t seem to be the case. This is probably because natural selection 4:27 preferentially made use of large molecules which reliably fold the same way. 4:33 There are some exceptions to this. For example prions, like the ones that are 4:39 responsible for mad cow disease, have several stable shapes. And proteins can change shape if 4:46 their environment changes, for instance when they encounter certain substances 4:51 inside a cell. But mostly, the amino acid sequence determines the shape of the protein. 4:57 So, the protein folding problem comes down to the question: 5:01 If you have the amino-acid sequence, can you tell me what’s the most stable shape? How to try and solve it 5:07 How would one go about solving this problem? There are basically two ways. 5:12 One is that you can try to come up with a model for why proteins fold one way and not another. 5:19 You probably won’t be surprised to hear that I had quite a few physicist 5:23 friends who tried their hands at this. In physics we call that a “top down” approach. 5:29 The other thing you can do is what we call a “bottom up” approach. This means you observe 5:34 how a large number of proteins fold and hope to extract regularities from this. 5:40 Either way, to get anywhere with protein folding you first of all need examples of how folded 5:46 proteins look like. One of the most important methods for this is X-ray crystallography. For 5:51 this, one fires beams of X-rays at crystallized proteins and measures how the rays scatter off. 5:58 The resulting pattern depends on the position of the different atoms in the molecule, 6:03 from which one can then infer the three-dimensional shape of the protein. 6:08 Unfortunately, some proteins take months or even years to crystallize. But a new method 6:14 has recently much improved the situation by using electron microscopy on deep-frozen 6:20 proteins. This so-called Cryo-electron microscopy gives much better resolution. The CASP Competition 6:27 In 1994, to keep track of progress in protein folding predictions, researchers 6:33 founded an initiative called the Critical Assessment of Protein Structure Prediction, 6:38 CASP for short. CASP is a competition among different research teams which 6:44 try to predict how proteins fold. The teams are given a set of amino acid 6:50 sequences and have to submit which shape they think the protein will fold into. 6:55 This competition takes place every two years. It uses protein structures that 7:00 were just experimentally measured, but have not yet been published, so the competing teams 7:07 don’t know the right answer. The predictions are then compared with the real shape of the protein, 7:13 and get a score depending on how well they match. This method for comparing the predicted with the 7:20 actual three-dimensional shape is called a Global Distance Test, 7:25 and it’s a percentage. 0% is a total failure, 100% is the high score. In the end, 7:32 each team gets a complete score that is the average over all their prediction scores. 7:37 For the first 20 years, progress in the CASP competition was slow. Then, researchers began 7:44 putting artificial intelligence on the task. Indeed, in last year’s competition, about half 7:51 of the teams used artificial intelligence or, more specifically, deep learning. Deep learning uses 7:57 neural networks. It is software that is trained on large sets of data and learns recognize patterns 8:04 which it then extrapolates from. I explained this in more detail in an earlier video. Alphafold 2 8:10 Until some years ago, no one in the CASP competition scored more than 40%. 8:16 But in the last two installments of the competition, one team has 8:21 reached remarkable scores. This is DeepMind, a British Company that was acquired by Google in 8:29 twenty-fourteen. It’s the same company which is also behind the computer program AlphaGo, 8:35 that in twenty-fifteen was first to beat a professional Go player. 8:40 DeepMind’s program for protein folding is called AlphaFold. In twenty-eighteen, AlphaFold got a 8:47 score of almost 60% in the CASP competition, and in 2020, the update AlphaFold2 reached almost 90%. 8:57 The news made big headlines some months ago. 9:00 Indeed, many news outlets claimed that AlphaFold2 solved the protein folding problem. But did it? What does this mean? 9:08 Critics have pointed out that 90% is still a significant failure rate and that some of the most 9:15 interesting cases are the ones for which AlphaFold2 did not do well, 9:20 such as complexes of proteins, called oligomers, in which several amino acids are interacting. 9:27 There is also the general problem with artificial intelligences, 9:31 which is that they can only learn to extract patterns from data which they’ve been trained on. 9:37 This means the data has to exist in the first place. If there are entirely new 9:42 functions that don’t make an appearance in the data set, they may remain undiscovered. 9:48 But well. I sense a certain grumpiness here of people who are afraid they’ll be rendered 9:53 obsolete by software. It’s certainly true that the AlphaFold’s 2020 success 9:59 won’t be the end of the story. Much needs to be done, and of course one still needs data, 10:05 meaning measurements, to train artificial intelligence on. 10:09 Still I think this is a remarkable achievement and amazing progress. It means that, in the future, 10:16 protein folding predictions by artificially intelligent software may save scientists much 10:22 time-consuming and expensive experiments. This could help researchers to develop proteins that 10:28 have specific functions. Some that are on the wish-list, for example, are proteins to stimulate 10:35 the immune system to fight cancer, a universal flu vaccine, or proteins that breaking down plastics. Sponsor Message 10:43 This video was sponsored by NordVPN which is a software that you install on your laptop or phone 10:48 and it keeps you safe as you browse the internet. You use the NordVPN app to connect to one of their 10:55 servers and browse the web from there. This keeps your data safe even on a public wireless. 11:01 Using this app has the added benefit that you can choose your virtual 11:05 “location” from any one of their more than five thousand servers all over the world. So, 11:11 if you ever encounter a video or website that is blocked where you are, 11:16 you can just connect to a server in a different country and access the website from there. 11:22 In a special offer for viewers of this channel, you can now get a 2 year plan 11:27 plus one additional month with a huge discount, 11:30 if you use the link nordvpn dot com slash Sabine and the coupon code sabine. That’s S A B I N E. 11:40 NordVPN works on pretty much all platforms, android, windows, iOS, 11:44 what have you. Once again, that’s nordvpn dot com slash sabine and the coupon code sabine. Sabine Hossenfelder 1.42M subscribers Videos About Patreon Twitter Instagram Facebook 6:47 10 differences between artificial intelligence and human intelligence by Sabine Hossenfelder 17:48 All you need to know to understand 5G by Sabine Hossenfelder Shop the Sabine Hossenfelder store Science without the gobbledygook Comfort Tee $25.99 Spring The Shocked Professor Mug $15.99 Spring This Guy Again Classic Crew Neck T-Shirt $19.99 Spring Strange Attractor Unisex Classic Pullover Hoodie $41.99 Spring This Guy Again Sticker $5.99 Spring 1,569 Comments rongmaw lin Add a comment... Pinned by Sabine Hossenfelder @SabineHossenfelder 3 years ago NordVPN special deal: Just go to https://nordvpn.com/sabine and use our coupon code sabine at checkout. 119 Reply 55 replies @epemsley3787 3 years ago My headphone cord thinks it's a protein...very annoying. 323 Reply 17 replies @TIO540S1 3 years ago Refreshing to hear someone use “exponentially” when it actually applies. 600 Reply 26 replies @jrherita 3 years ago Sabine - really appreciate you taking the time to explain these complex topics to us. Thank you! 386 Reply Sabine Hossenfelder · 3 replies @SangsungMeansToCome 3 years ago The silver rule of broadcasting: Whenever there is a question in a headline, the answer is always NO. 201 Reply 12 replies @mg4361 3 years ago As a biochemist who worked both with x-ray crystallography and cryo-EM, a few comments: a) this is an excellent video, but i also never expected anything other than awesome from Sabine b) even if alpha fold performs less well for complexes or unusual folds, still it would be great if it could releive the burden of having to determine the structures of hundreds upon hundreds of similar proteins. This would free up valuable people and equipment to concentrate on the very unusual and complicated structures c) even if it doesn't revolutionize basic biochemistry, it can be of tremendous help to pharmaceutical research as they often work on relatively common folds (e.g. spike protein of Covid when you already know the spike structures of several other coronaviruses) 361 Reply 45 replies @zeawoas 3 years ago (edited) A few clarifications: CryoEM doesn't generally give better resolution (it was actually worse than crystallography until recently), but it also works on many proteins that are not amenable to crystallization (e.g. membrane proteins). That's why it's so useful. Also, oligomers are not several amino acids interacting, but several distinct amino acid chains (i.e. proteins). 316 Reply Sabine Hossenfelder · 33 replies @illogicmath 3 years ago Sabine's social work in disseminating knowledge on such a wide range of scientific topics is invaluable. 70 Reply 6 replies @RichMitch 3 years ago Congratulations on 200k, much deserved 1 Reply @johnhoebel8209 3 years ago Excellent. I learned so much in 12 minutes about a subject I was totally ignorant of. 66 Reply @subhrodeepsaha9245 3 years ago I just started my PhD in peptide folding and chemistry and then comes this news!! 85 Reply 22 replies @MFJM 2 years ago As a Senior Biochemistry student in 1985 I wrote a computer program which combined Chou-Fasman and Robson-Garnier's methods of predicting Secondary structure. Validation would require a huge amount of reruns on parameters based on known 2nd and 3tr structures, however, only a few structures of Albumin were known at the time. Reply @jdkingsley6543 3 years ago The fact so much of these problems were being tackled in the 90s and before is mind blowing. The possibilities of a collective mind 2 Reply @datapro007 3 years ago Hi Sabine, I'm echoing what a lot of other folks have already stated. I really enjoy the diversity of topics that you present, and your well thought out presentation too. Thank you. 30 Reply @davidhand9721 3 years ago The problem with AI is that even when it understands something, it can't teach us anything about what it's learned. 23 Reply 6 replies @michaelblacktree 3 years ago I used to do Folding@Home, back in the day. It felt good to help science, even in such a small way. 2 Reply 1 reply @mattisalmela1734 3 years ago Protein structures can also be studied using NMR. This has the advantage of not having to crystallize proteins and having them in their "native" form in solution. Great video as usual! 10 Reply 1 reply @asafnisan 3 years ago You deserve a lot of respect for exposing us to different domains and questions in science. In a very understandable way too. 8 Reply @BryanWLepore 3 years ago CASP updated the scoring algorithm in 2014 and the top scores have increased linearly since then. The C-alpha “IDDT” distance scoring is based only on C-alpha positions - that is one atom out of all other amino acid atomic positions. There are other scoring calculations too. 16 Reply 4 replies @Spock0987 3 years ago (edited) Botton down, botton up and not anywhere closer to solving this problem. I have always been worried about the folding problem myself, since I first heard of the problem in elementary school. Anyway hope we are almost there and can solve this mystery before I departure. Fantastic lecture and I appreciated every minute. Reply @hedwegg 3 years ago Topic: Protein Folding & Deep Mind. A Great Journey ahead. 1. To the Quik: [Contour, Curvature & Rhythm] are the major elements for [Folding]! 2. [Deep Mind" "hasn't been programmed" [to recognize patterns & outcomes] using [Contour, Curvature & Rhythm] for a [Protein] to [Structurally Fold] into a [Functional & Recognized Arrangement]. i.e. a [Living Organism]! Great job by the Commentator! Looking forward to the present [Robotic Era]. (Boston Dynamics)! 1 Reply @davidschroeder3272 3 years ago This was just an absolutely fascinating topic. I had heard of this problem years ago, but never bothered looking into it. Sabine's ability to unravel the complexity of such issues is remarkable. 4 Reply @SWatchik 3 years ago Great explanation and nice seeing more Biology content on here, as a Biologist myself! 71 Reply Sabine Hossenfelder · 7 replies @senthilkumarpalanisamy365 3 years ago Thanks for the clear cut explanation, Sabine 1 Reply @lucaspierce3328 3 years ago (edited) Protein Folding is one Subject I love to study in Biophysics! Although I haven't been doing to much research on it lately! But later this year I will start Again! An Important element of Protein Folding involves the Main Colloidal Medium known as Water more Specifically Liquid Crystalline/Quasi-crystalline Water, Water Exclusion Zones, & Related! Liquid Crystalline/Quasi-crystalline Water is a Quantum Super-Colloid(it also can have Superfluidity-Quantum Biochemistry) which acts as a very effective Ratchet of Brownian Motion which decreases Randomness and Increases order being Syn-entropic producing more Regularized, Organized, & Stably Repeatable, & Robust Arrangements of atoms in a Macromolecule like Proteins(for example Environmental Parameters like PH, Salinity, Temperature[HSP & Chaperonins so on] & others can effect the Folding Process)! Protein Folding has Evolved to reduce much variation or evolved Robustness(Differential Inter-&-Intra-Modular[Topo-Modularity] Plasticity and Robustness)! Although there's always going to be Stochastic Variability and Structural Plasticity in Protein Folding!. Reply @MarianneExJohnson 3 years ago "That's almost as many vacua as there are in String Theory" -- priceless 146 Reply 16 replies @henktl3580 3 years ago Good string theory burn! 38 Reply @AhmedAshraf-pd7mu 2 years ago A biologist here, 6:21 Cryo-EM doesn't give better resolution. Actually, relatively low resolution is one of the drawbacks of Cryo-EM. In the past, resolutions were very bad, but in the past few years, great development in algorithms used to 3D reconstitute the structure (the electron microscope gives you 2D projections of the structure, and you need to reconstitute the 3D structure out of these projections) and developments in the electron microscopes themselves have improved resolutions a great deal and made cryo-EM a very attractive approach (personally I believe along with predictive AI software, Cryo-EM is the future of structural biology). On the other hand, the advantage of CryoEM is that it is much better at imaging large protein complex compared to X-Ray chlystarograhpy. You don't need to crystalize the protein (which is a pain in the ass), it is better at capturing biologically relevant structures because the proteins are in solution (before being flash-frozen) compared to the crystallized proteins, and Cryo-EM is much better at capturing protein dynamics (which is very relevant to the function) 1 Reply @LoisSharbel 3 years ago Thank you for clarifying this difficult information for lay people like me. Yours is a vital service and you are so good at it. Getting the general public to 'somewhat' understand the amazing discoveries coming so fast now is urgent. Reply @pilover314159 3 years ago I really appreciate how you out the science back into popular science, you make real science popular, not that phony science that is found in self help books or on certain pod-casts that shall not be names 3 Reply 1 reply @paolo7733 3 years ago "That's as many as there are vacua in string theory" do we live in a world where string theory is more widely understood than protein folding? 175 Reply 23 replies @Tom55data 2 years ago Methods for protein crystallography, and molecular simulations, was my field for many years. I would like to add to your nice explanation why even with the good answers ,deep mind is not quite there; and it comes down to the precision of the answer, and the accuracy to the true answer. If we take catalysts, much of the protein is a scaffold, a frame work to hold a few atoms in the right orientation, but with enough flexibility to enable a reaction to occur. To understand these reactions and how the proteins work it is necessary that the atom positions are well determined, or we cannot tell if a reaction is SN2 or Elimination - and without that knowledge we cannot determine how that protein fits into the metabolism. The the fundamental process of solving the structure of a protein is therefore to discover the function of the protein and we can only do that if the atom positions are well enough defined, and the flexibility can be observed that allows understanding. Therefore, protein folding must be good enough to discover that, and a score of 90% for the whole protein (or group of proteins) says nothing about the small number of critical atoms. That problem will be solved at some stage, but I would say, not quite yet, as we still cannot replace the complex process of structure determination for folding programs even for quite simple structures. 1 Reply @GagandeepSingh-rz7ue 3 years ago You are one of the best and honest youtuber. No hype, to the point explanations. Reply @janerussell3472 3 years ago Proteins need chaperones to fold, the correct enzymes. For example, for human mitochondria, after synthesis on cytosolic ribosomes, nuclearly encoded mitochondrial precursors are imported into the organelle in their unfolded state ( due to the limited pore size of the import machinery. ) After import, mitochondrial precursors must efficiently fold into their functional structure, to avoid aggregation due to exposure of hydrophobic surfaces. In the mitochondrial matrix, [ human mitochondria contain their own genome encoding 13 core subunits of the oxidative phosphorylation (OXPHOS) machinery ] two highly conserved chaperone systems, HSP60 and mitochondrial HSP70 (mtHSP70), are critical in facilitating the folding reaction of mitochondrial precursors. Proteins don't fold in isolation of their environment or intended functions. Sorry to invoke intentionality. 5 Reply 2 replies @arik9112 3 years ago ''some of you may have folded proteins yourself'' YES I HAVE WeLL IM ALIVE, that's why!~ 115 Reply 5 replies @GiuseppePaleologo 3 years ago A family member is working primarily in drug design and protein folding. In his view, this presentation was accurate and balanced in its conclusions. I did not know Hossenfelder was doing scientific divulgation; providing a competent, informative and non-hyped account of scientific news is a very valuable service. Reply @soulsdegreen 3 years ago Complex indeed...even before mentioning the huge amount of post translational modifications like glycosylations, lipidation and phosphorylation...which all affect how AA interact with each other but also changes surface binding properties enough to change association with other biomolecules in a cell... 1 Reply @johnr.5475 3 years ago Thanks. I really do enjoy my weekly fix of interesting stuff. I was actually wondering how they could test that claim of solving the protein folding problem - now I know. 5 Reply @robinhodson9890 1 year ago A few months after this video was made, Alphafold made a little more progress, becoming usefully accurate for the easy cases. Although far from solving the overall problem, this still saved several man-centuries over older cataloging efforts, which could lead to much faster drug discoveries, and shorter clinical trials. 2 Reply @augustaseptemberova5664 2 years ago In context of "human researchers becoming obsolete" I'd also add that Deep Learning acquires implicit logic, meaning you can make it make great predictions and hypothesis, but you can't inquire what principles, patterns etc. DL learned from the training data that allowed that prediction, the how or why DL builds its output. You'll always need humans to make sense of DL. (and as said in the video: humans to provide and select good training data etc.) 1 Reply @BinaryReader 3 years ago I've seen a few of your videos in passing before, they're always very good. Subscribed today :) Reply @kamalnawhal9356 3 years ago I just love the way you explain all these things and I easy that sounds thanks 7 Reply @qbasic16 3 years ago Wonderful video, Sabine! 5 Reply Sabine Hossenfelder · 1 reply @v.567 3 years ago Αs someone who wants to go into Immunology-Microbiology, Toxicology, Neuroendocrinology and Cell Biology, people in Computational Biology, Protein Crystallography and Phycisists who work on stuff like that are intimidatingly smart. Lots of respect for every Scientist out there and all the Healthcare workers giving all they have to fight against the Pandemic. 1 Reply @dankuchar6821 3 years ago Love the subtle String Theory dig. Good job! 1 Reply @SteveHill3D 3 years ago Superb as always. My lay-person understand is significantly enhanced! I am thinking that the folding of engineered proteins might still be much harder to predict as they will not have run the gauntlet of natural selection. 3 Reply 1 reply @baharuddinbukari3088 3 years ago .. I like and appreciate your infomative, matter-of-fact, neutral views. 13 Reply @hojoj.1974 3 years ago I first came for the music, then stay for the science. Doctor you are amazing at explaining things without the gobbledygook. And I thank you for that. 2 Reply @joedart1465 3 years ago This is a very, very well done video. The protein folding problem was put into perspective without over or under dramatization. Excellent delivery. Reply @paulsmall5122 3 years ago Interesting and clearly explained as always. Many thanks. 4 Reply @ahmedsalman9906 3 years ago Great video! I would love to see more Biology and Chemistry content here. 4 Reply @LowellBoggs 3 years ago Another great video - I appreciate the wide range of topics that you explain in your videos. Thanks 1 Reply @ilya4759 3 years ago I don't know much about protein folds, but I know a lot about fat folds, and how to predict their shapes. Will that win me a Nobel prize? 1 Reply @rightright6582 3 years ago (edited) She knows how to stimulate our minds. Keep it up please. 8 Reply 1 reply @johnathondavis5208 3 years ago How wonderful it would be if Sabine were my next door neighbor....with a LOT of patience for questions :) She is just brilliant and so enjoyable to listen to. 12 Reply 1 reply @philjed5178 3 years ago the String Theory Burn was SICK :D 1 Reply @siddharthb2633 3 years ago As an engineer, I am programmed to asked," How is it useful?". But your channel has really helped me appreciate pure sciences . Thank you. Reply @nzuckman 3 years ago I had a project at NIST where my job was to write a program that could model how proteins plump up in an aqueous solution! Maybe it finally can be put real use haha! 6 Reply @ashertoh7097 3 years ago Sabine is really fantastic. She is the best. 4 Reply @jeremyredd4232 3 years ago One of the comments below was super interesting. Paolo Privano asked, "do we live in a world where string theory is more widely understood than protein folding?" I wonder how you would illustrate the difference between something we know exactly how to solve with brute force, even though that method of solving it is intractable, and a problem with no known solution, even though there are good candidates. I also wonder how a beyond standard model physicist would explain the following to a lay person (i.e. me) and why these are such difficult problems: A) Why there are several candidates for the group structure of string theory, SU(5) vs SO(10) or somehow even the monster group gets mentioned. If it's the monster group, that's a lot of symmetries to preserve. B) Why even the most common string theory group structure U(1)xSU(2)xSU(3)->SU(5) results in ~10^250 potential algebras. I mean I you are mapping multiple surfaces onto a surface and finding that tangent space. I get that there is more than one way to do that, but 10^250 ways to do that? Reply @halporter9 3 years ago Excellent, succinct, communicates a complex understanding of progress on a very stubborn problem. Reminds me of the Imagination and sweat that led to the description of the structure of the heme in the 1950s. Reply @DJVARAO 3 years ago The Ai team just exposed how inefficient the other computational techniques have been. Molecular dynamics in particular, is such an inflated but inaccurate tool, that maybe is time to start thinking about the seriousness of "bottom up" approaches. Starting by the fact that you cannot simulate non-covalent interactions at the required accuracy even with quantum mechanics for big systems just yet. Much less with Newtonian mechanics, as 99% people in this field foolishly do. But they get grants, so why they will gonna change that? Out of embarrassment? Certainly not. 7 Reply 24 replies @BryanWLepore 3 years ago (edited) The “protein folding” problem(s) are one thing - what appears to be shown here is a “protein folded” problem - various protein constructs folded in a plethora of expression systems and crystallized or otherwise isolated in states conducive to structure determination - notably not including NMR methods - and deposited in the RCSB. That’s a magnificent endeavor but the programs are not showing how or what the protein went through to fold - it is showing the likely final result in the RCSB except for NMR structures. I think fold space is filling up, so this might indicate the methods to determine structures are not discovering highly unique structures at the moment. Is the work peer-reviewed yet? 6 Reply 7 replies @ninomeloni9671 3 years ago Amazing !Clarity and comprehensiveness in this brilliant video.Thanks! Reply @ThomasJr 2 years ago When it comes to Science, Zabina is the best teacher out there for lay audiences. No kidding, truly. She's accessible Reply @reidmock2165 3 years ago "That's as many as there are vacua in string theory" And this is why I love Sabine 3 Reply 6 replies @mheermance 3 years ago Then unnerving thing about AIs is that they mostly work, but they always have edge cases where they fail, and don't know they've failed. In fact they're often quite confident about their erroneous prediction, and fixing them is difficult. I imagine optical illusions are this phenomenon in humans, but we at least understand we're wrong. 3 Reply 10 replies @meows_and_woof 3 years ago Love your videos! I’m a biomedical engineering student and I find your channel really informative. You have become one of my favourite content creators Reply @DigitalAlligator 3 years ago (edited) 3:35 love your brute honest mocking 1 Reply @rmehta54 3 years ago Very interesting! A couple of question: what makes a fold stable? How can software determine if a fold is stable or not? 3 Reply 4 replies @eljcd 3 years ago Protein folding? Sound interesting, thanks, but that remind me I need to do the laundry... 4 Reply @gaborendredi8161 3 years ago I love your sense of humor. I laughed loudly when out of the blue you said ten to the three-hundred possibilities, that’s almost as many as vacua there are in string theory. How do you decide your dress and hair style? You are fantastic in this video! 1 Reply @dinoj1734 3 years ago The problem is way more complex than just the 'folding' of the 'protein' 'string'! (Pun intended) Firstly you have to have the correct protein folded into the correct shape in the correct place in relation to other proteins folded into their correct shape all at the correct time. So even if you have proteins that 'usually' fold into a particular shape (we will just ignore the ones that don't) you still have to have the correct ones in the correct place at the correct time... This goes way beyond exponential, way into the territory of statistically impossible!!! 2 Reply 1 reply @curious_atoms 3 years ago As a former combichem drug researcher, I must admit, this makes me want to be more rational. 9 Reply 5 replies @sadface7457 3 years ago (edited) It has not been solved. This is just another tool for numerical optimization. 7 Reply Sabine Hossenfelder · 4 replies @GururajBN 3 years ago You never cease to amaze me. So far I thought that you were a rebellious theoretical physicist occasionally singing pop music and uttering catchy quips. Now you come out to explain an intricate problem of biochemistry. Awaiting more such pleasant surprises. Reply @paulsaunders7945 2 years ago An excellent review of recent advances in AI application to solving protein folding problems. Although a great advance in protein structure prediction, it works for the subset of proteins which are readily crystalizeable and the primary determinant is the amino acid sequence sans post-translational modifications. There are a lot of proteins that are difficult to crystalize, such as membrane proteins, components of multi-subunit complexes and glycoproteins. A major area of protein folding research will continue to be how chaperonin proteins such as proline isomerases and heat shock proteins (HSP) overcome thermodynamic barriers to guide proper protein folding. 1 Reply @d4v0r_x 3 years ago if by "solved" you mean "not solved", then yes, its' been "solved" 13 Reply 1 reply @anonymousperson2640 3 years ago (edited) 9:28 That is actually called machine learning, not artificial intelligence. And it's not entirely correct that it can only learn on existing data, AlphaGo learned without any data, governed by a set of rules which helped to determine whether one option was more preferable than the other. From evaluating those rules on sets of data from initial runs with a random values, the machine learned which one values are better for the task. That's is probably just how it learned folding too, since it's pretty easy to determine which resulting fold is better than the other... 7 Reply 6 replies @Poopyduckling9999 2 years ago The Lavinthal's paradox literally blew my mind when I was in undergraduate. Reply @brucecheesman2781 3 years ago (edited) An excellent video. Just one small point:- multidimensional NMR spectroscopy is also used to determine protein folding, in addition to X-ray crystallography. The NMR spectroscopy approach makes use of internuclear distances. Reply @gio5969 3 years ago These videos are extremely disturbing. Intelligent, rational explanations in YouTube videos? Talk about crushing my world view. 28 Reply 3 replies @kevfquinn 3 years ago aside; I find the high number of shot changes (almost every sentence, and hopping left, right, forwards, backwards) to be quite distracting. Fewer would be easier to follow, I think. 3 Reply @daveinwla6360 1 year ago Predicting a folding configuration should start with its assembly method - amino acid-by-amino acid - and predict the folding as the chain incrementally lengthens. That would greatly reduce the number of possible configurations, the lowest energy state chosen at each point of the assembly rather than the whole chain at once. Reply @antoniomaglione4101 3 years ago Thank you for producing this video Dr. Hossenfelder. Duly liked and shared. I've been looking into the field a decade ago, before the attempts with AI. We start with a protein codified by the linear, 2-D code of the DNA. As the ribosomes in the cell finish to assemble the protein, as soon this sequence of linked aminoacids leave the cell, within few milliseconds it folds into the required shape. This happen due to the complex interactions of electric fields the molecule possesses. My (informal) research consisted in locating, within the molecule, a number of electric dipoles, and elaborating a matrix which completely define the molecule. Each dipole is a 3-D vector, and the matrix gets quite big quite fast; I just assumed this isn't a problem if a computer with the right power is used for the calculations. Interesting how harmonic patterns emerges from the molecule electric field. You rightly pointed out that AI require some initial pattern definitions in order to conduct a recognition. This is why AI wouldn't initially work in the field of proteomics. But that kind of AI is what Google excels at: picture it, find pattern in the picture. I believe is not going to work well. They don't need X-ray crystallography or electron microscopy, but a nanometer-sized array of Hall sensors, so they can map the dipoles by moving the molecule within the Hall's sensors array. Regards, 1 Reply @cmpe43 3 years ago I dont think folding should be the action term used for proteins because when I fold my clothes this way, i always get into trouble with the wife. 3 Reply @gregrice1354 2 years ago Thanks again Dr. H.! You are helping build critical context for public grasp of important and relevant science, industry and social action. Reply @daffidavit 3 years ago (edited) I'm a follower of your channel for things that have to do more with physics, than chemistry. But, attempting to be a good student I've listened to this lecture. Honestly, I could not follow it very well, not because of Sabine's lecture, but due to the subject matter. I studied chemistry in high school in the U.S. and college at FIT in Florida in the early 70s, but my grades in chemistry were only average. I left engineering school after two years to obtain a BS in business administration. I later went on to obtain a doctorate degree in law from a university near Chicago. But my heart has always been in science. I love Sabine's channel but I find chemistry to be very difficult to learn. It was hard for me to learn as a kid and even more difficult as I get older. But I have a proclivity to Sabine's lectures in particle physics and astronomy for some unknown reason. I can't explain it, but that's just the way my brain works. How is it I can Sabine's lectures in quantum physics but have no clue when it comes to chemistry or biology? Reply @beckmack1994 3 years ago (edited) I'm more confused about the protein folds than before this video. You bring the supposed problem of protein folding, not exactly, and then you failed to explain it by confusing it further. I don't think you full grasp what the issue really is. Or, you failed to explain it 3 Reply @lajosbaranyi7333 3 years ago (edited) It is important to know: there are 20expN possible protein sequences of N lenght. Out of that only a super minuscule number can actually fold and has definite native structure. The rest are in an non - definite state, called molten globule. The AI can not do anything with 99.9999999999999...999999.9999 percent of “proteins”. Only a fraction of NATURALLY occurring proteins can be tackled by the algorithm. However the AI currently cannot tell you wether a particular protein is naturally occurring or pulled out of a magic hat. It does not look like solving the problem of protein folding. I do admit though that the AI is good in pattern recognition as long as there are real patterns. Current state of affairs is not unlike looking for your list key at midnight. We are only searching under the street lights. Mist importantly though I thoroughly enjoyed Sabine’s very accurate and eye opening explanation . 2 Reply 1 reply @h.i.5280 3 years ago This kind of channel is the best use of the internet. Thank you Dr Sabine! Reply @727Phoenix 3 years ago I clicked on your song Outer Space right after your lucid explanation about protein folding. Imagining proteins fold while watching you dance created weird mental images ;-) Anyways thank you for teaching us this important topic. May you and your family have a great weekend! Reply @deepz513 3 years ago Very insightful..the way u explained is endearing Reply @michaelworkman4057 3 years ago Thank you Sabine for another wonderful video, also for mentioning the need for a protein that breaks down plastics. Any help with that is needed right now. Reply @raghu45 3 years ago Thanks again! The topics you choose are very interesting and you cover it with quality erudition. But please, please, increase the recorded sound volume 2 more steps. Thanks. 1 Reply @36nibs 2 years ago I was playing Fold it in highschool Legit randomly remembered and watched this to catch my memory up! Reply @CristianKlein 3 years ago "And ... ahm ... what was the protein folding problem again". :)) No minute is wasted when watching Sabine. The perfect balance of entertainment, scientific rigor and "Sabine sarcasm"(tm). 1 Reply @airmanfair 3 years ago Such a great video. The purity of the educational content is so high. Reply @IBITZEE 3 years ago Finally... a 'construted/solid' explantion of what/why/how proteins work... now waiting for a video on how molecules bind... what angles are made... what's the 3D result... (the folding process) also a video about prions would be welcome... ;-) Reply @sanjitdaniel4588 3 years ago Awesome Sabine! Answered a lot of my questions. Thanks! Reply @thejamaican67 3 years ago I love your channel you explain complex concepts and making them understandable Reply @yimkwonglun62 3 years ago You really open my horizon, I am deeply appreciated your teaching. Reply @blueberrylane8340 3 years ago Such a good presenter, few people combine both knowledge and no-nonsense charisma like this. Its refreshing to get an outright take. Reply @davidwilkie9551 3 years ago And that lecture is what makes (in the Eternity-now Interval Conception of self-defining temporal substantion.., of time-timing making), Physics, mathematically dominant Sciencing, of how the biological clocking of modulating probability in Quantum Computational Chemistry, is why accurate evaluation of data, and precise Theoretical Modelling, is the reason to study Physics in conjunction with other subjects. Reply @Val-nq5pt 2 years ago This is a helpful intro to protein folding. Thanks Sabine! Reply @vblaas246 3 years ago Molecular Life Scientist here (MSc), drug design needs a known protein structure but that is not all. Proteins can be modelled as hard spheres obeying Newtonian physics, but predicting drug binding is inherently a N-body problem where electron-electron interactions need to be taken into account too. We have Car-Parinello QM/MM simulations, but that is still far away from being able to predict drug binding. Besides side-effects appear mostly due to drugs binding other proteins which have nothing to do with illness. Reply @robertschlesinger1342 3 years ago Interesting and worthwhile video. Many thanks for the link to a relevant paper. Reply @paulwallis7586 3 years ago Very clear explanation of such a tricky subject. Meanwhile - Hm. Occam's Razor - You know the probability of recognized stable forms, therefore you can discount sequences which typically don't deliver stable forms for prediction? AI could learn from predictive data sets which types of predictions don't work on a statistical basis. Reply @elontusk610 3 years ago 5:50 I had no idea. Thank you for sharing! 2 Reply @robertgotschall1246 3 years ago Been getting this in bits and pieces. Thanks for bringing it together for me. Reply @GG-dx6cu 3 years ago Thank you again for a great and stimulating discussion. One pretty important thing should not be forgotten: one key factor for protein folding (and also DNA structural changes B<—>Z) is the interaction of the protein (or DNA) with its solution (H2O networks stabilizing structure around polar protein groups) and of course the local concentration of ions (Na Cl). I remember that the crystallization of lysozyme happened in a small window of NaCl concentration and not below or above. A group at the MPI of BPC modeled those interactions with changes to the free energy W(1,2) between two particles 1 and 2 based on the changes to ion concentration in the “background”. Reply @alandoran9891 3 years ago Very good introduction to a fascinating topic 1 Reply @BigZebraCom 3 years ago Thanks Dr Sabine, I enjoyed this video! 1 Reply @nearlyzero9849 3 years ago The brief string theory blip at 3:30 was hilarious! I laughed out loud. Nice one Sabine, loved it. Reply @KeithRowley418 3 years ago Excellent explanation of the problem and potential solutions. Thank you. Reply @ascaniosobrero 3 years ago More than design new specific proteins, at least in the pharma research world it is crucial to know or predict the 3D structure of proteins (enzymes, receptors, ...) to design new drugs. Small molecules that bind effectively the protein, for instance in the active site of an enzyme to block its function, or to antagonize its activity. Not sure this procedure is effective enough with membrane receptors, which are usually extremely difficult to crystallize out of their environment, and often work forming dimers or trimers or higher structures. 1 Reply @roros2512 3 years ago finally I understood this problem, thanks for the nice explanation 2 Reply @Viperzka 3 years ago This pointed out what I think is the best use for the tool. Scientists run the system to find a theoretical protien that does what they want, then they run an experiment to see if the protein does in fact work. Reply @seanspartan2023 3 years ago This is the first time I am hearing of this. Thank you, Sabine 2 Reply @vineelreddypandi3861 3 years ago Thank you for explaining such complex problem so clearly. Reply @andrewpaulhart 3 years ago Thank you . By far the best discussion of this subject that I have seen Reply @dylanmah7791 3 years ago Just a small correction from a molecular biologist on one of your points: CryoEM does not have superior resolution to X-ray crystallography as a technique for probing protein structure, it’s generally poorer in comparison. The advantage of CryoEM is that you don’t need to crystallize your proteins, which isn’t feasible for many proteins, and unlike solution NMR, CryoEM doesn’t suffer from a the same kinds of limitations on the size of your proteins. Reply @bozo5632 3 years ago Thank you for the videos, I will never get enough of them. Reply @slowhandsoff 3 years ago (edited) Wonderfully fascinating, molecules, evolution, physics and some measure imperfections over Spacetime! Thank you for what you do and share in your videos! My faint understanding is always very much more illuminated during and ever after such encounters. Reply @matthewdunstone4431 3 years ago I have just discovered your channel. Fantastic content. Thank you so much. Reply @patrickmchargue7122 3 years ago A great explanation of the problem. Thank you. 1 Reply @sir-yz7cw 3 years ago You're the best, Sabine! Keep up the great work! Thank you. 1 Reply @grikalanovv 1 year ago The problem isn't calculating the probability, but positioning the amino acids for actual synthesis. Or you can make a "soup" counting on better probability, but it's not controlled synthesis. Reply @Panterakat1 3 years ago This was actually quite fascinating. Thank you Sabine Reply @chemistproffatima 3 years ago (edited) Hello sabine, i found this video really useful to me as a 13 year old girl.... Since I'm interested in organic chemistry and biochemistry! Thx again :) plz upload more biochemistry videos I'll watch them 1 Reply @anonym0usplatypus 3 years ago As a researcher in an unrelated field, it is refreshing to hear a realistic report on the state of science. Thanks Reply @someguyintexas1066 3 years ago Great presentation, Sabine! I’ve read that some say that the first pre-life on Earth could have been a lot like modern prions, just protein molecules with the ability to fold other molecules into copies of themselves. I knew Mad Cow Disease was caused by a prion, but it didn’t know it had multiple stable configurations. Learn something new every day! Reply 1 reply @whiteboar3232 3 years ago Another problem with neural networks is that knowledge embedded in a neural network is not readable, understandable. 1 Reply 1 reply @aasemal-lmki8286 3 years ago Wow. I just love this explanation. Thank you Reply @Trig188 3 years ago Sabine - love your videos. You do a great job of explaining the complex. Besides protein folding, a video on contoured bedsheet folding would be nice. ;-) Reply @Katherine-L789 3 years ago I always enjoy these videos! Reply @xyzct 3 years ago Like soap films across wire frames, proteins folding is a great example of an analog computer. Reply @bhangrafan4480 3 years ago This has been a major preoccupation of mine since being an undergraduate in Biochemistry. It is interesting to hear about these developments, but only time will tell how much success has really been achieved. The problem is in fact to predict the 3-D structure from the amino acid sequence which can easily be obtained from genetic information. Actually determining the 3-D structure directly is still a slow and difficult process, using techniques such as X-Ray diffraction or proton-NMR. Reply @kencf0618 2 years ago Good update on the protein folding problem. Reply @shelley-anneharrisberg7409 3 years ago Thanks so much - new next to nothing about this problem! What a great introduction - and so interesting! I can't get over how fascinating biology really is - unfortunately, they didn't make it so at my school! ;) Reply @FredMontier 2 years ago "... vacum in strings theory" tongue in... Well done !!! Reply @MyMy-tv7fd 2 years ago (edited) at 3:44 - the problem of folding proteins is not stability, it is functionality - eg, an enzyme has to be the right shape to catalyse the reaction of its substrate, which is not the shape of greatest stability. Stability is important for structural proteins. Reply @khosrofakhreddini7824 3 years ago Very good explained. Thanks Reply @jamieshelley6079 3 years ago (edited) Thanks for the great video! (As an AGI researcher )It's good progress, but it's very misleading but good to have a tool for comparing results, which is what the tool does. I'm working on solve the problem of learning from trained data only with my work, time will tell how successful it is.. at the very least it attempts to create understanding rather than just trans-formative correlation (DL). Reply @seanmulvihill3592 3 years ago Thanks for your videos. They have helped me improve my own general knowledge Reply @DMahalko 2 years ago I am no expert on this but folding is not entirely random because it does not come into existence instantly. It starts small and gradually becomes larger. So initially the total number of possible shapes is very small. Also new strings do not assemble in empty space with nothing around them. They are constantly being pushed around by the cytoplasm of the cell, and by other proteins in the vicinity of the new protein. The assembler proteins may themselves restrict the number of possible positions of the new chain as it self-assembles. Also what keeps the new protein string from sticking to the assembler protein which itself is already a protein string? How does it "know" to not attach to the assembler protein producing it? Why don't we end up with tangled knots of newly formed protein strings randomly attaching to other proteins around it in the cell? 1 Reply @davidderidder2667 3 years ago I am really learning from and enjoying your work Sabine. Reply @bonob0123 3 years ago 6:24 cryo-EM does NOT give better resolution than xray crystallography, it usually gives worse resolution but the tradeoff is it doesn't require the difficult process (many times impossible) of getting the proteins to crystallize. Reply @Weissenschenkel 3 years ago The stock market likes a lot when new things promise to solve old problems. For many years I donated computer power to projects like SETI@home and Folding@home, even knowing that it would be probably in vain. AI with deep learning is more of a brute force approach, so... it heavily relies on computer power. 1×10³⁰⁰ is a lot of iterations to be done (yet that number was new to my knowledge.) Thank you for clarifying! Reply @alamagordoingordo3047 3 years ago Sabine videos, Always the best quality. 1 Reply @daltonbode6909 3 years ago I feel like I just watched a very well written publication. I'm curious to know what an acceptable percentage would be and whether A.I. will be able to successfully guess the appropriate folds needed for all (or most of) the different forms of proteins. I'm excited to see what the future holds. Reply @cbbhvjc 3 years ago Nice discussion, thanks for posting. Reply @stevemartin4249 3 years ago (edited) I have a Quora buddy who works for Deep Mind ... and though I won't go into details here to protect his privacy, met him and his wife for an evening chat at Tokyo Japan's Google headquarters. We spent at least half of our time talking about personnel problems at Google Japan — the tendency for otherwise bright Japanese employees to aspire to a 'safe' position in the corporate hierarchy, and then continue playing it safe to protect their position, thus minimizing risk-taking and innovation. Anyone who has played the publish-or-perish game of tenure in academia will recognize this all too well. The above is merely institutional rigor-mortis, but it can deteriorate to enough corruption many institutions (mission drift) that quite a bit of research has been directed towards CWB (Counterproductive Work Behavior). I have personal interest in this because about 6 years ago, I resigned in protest from a tenured position at a Japanese College because of institutional corruption, micro-managed racism, and dark-triad behavior. But it can become even worse ... any A.I. enhancement of technological innovation can, and will, be weaponized. From Mary Shelley's original 'Frankenstein' to Michael Chrichton's 'Jurassic Park', thinkers have tried to warn us of the danger of fetishizing science and rationality itself ... have tried to sound the same warning that Robert J. Oppenheimer uttered on seeing the first atomic blast. We can not put the genie back into the bottle, and I am afraid it will not end well with us. I really like Sabine's videos — she is a great communicator — and have notifications turned on, but she seems more optimistic than myself that technical progress can be controlled by collective moral progress. Science as a heuristics may be the best hope we have. But as described above, is a double edged sword. Karl Popper is the go-to-guy for that warning. And along with T.S. Kuhn, reminds us that scientists work for the corporate nation state, and have their own mortgages to pay, personal agendas to follow, and can be as mired in petty politics as any Wall Street Hedge Fund player. Evolutionary biologist Ernst Mayr posited, as described by Noam Chomsky described in his 2010 Chapel Hill speech ... 'Human Intelligence and the Environment', that human intelligence may prove to be little more than a fatal mutation of a social primate, and he gave a good justification for that by pointing at the evolutionary arc of apex species. Stephen Hawking put it a bit blunter ... 'Stupidity and greed will mark the end of the human race.' I hope to see Sabina bring in a more moral framing of technical progress with references from the social sciences and liberal arts. But once we go down that road, it is difficult to impossible to separate assumptions underlying science from some similar political presumptions that are not particularly corporate friendly. We are already increasingly seeing the debate about this in determining whether social media such as ABC (Google/YouTube, Facebook, and Twitter) are platforms or publishers. But straightening this out will only be temporary. Even that will probably not be enough to save us from ourselves. 1 Reply @jcj2249 3 years ago Very pleasant to follow. Thanks. Reply @Killer-mx7tr 3 years ago I have one correction about AI, when you said "If the model was tested on data that it has not seen it will fail", each model has two data sets, one for training and one for testing, the "train" dataset is the one that will be used to train the model, the "test" dataset is the one used to test the trained model *it is a completely different dataset from the training dataset*, when you test the model on the "test" dataset you get an accuracy (number of correct predictions / N), and there is something called "epochs" that you can increase to increase the accuracy of the model (not to much so it wouldn't be overfitting on the train data). My point is you are right, if one category of those proteins were not included in the dataset then the model would not be accurate on predicting that category, therefore the problem is with the data and not with AI, so it is just a matter of time until we reach accuracies above 98% (aka more data for the model). 2 Reply 3 replies @walterblanc9708 3 years ago Thankyou Sabine, always nice to learn something :) Reply @hoaxuan7074 3 years ago You can also fold non-biological molecules, for a bigger range of things to do. Then do computational self assembly calculations to make self assembling robots. 1 Reply @marsupius 3 years ago Sometimes I prefer a top down approach, and sometimes I prefer a bottom up approach. It depends on who I'm with. Reply @luiggiphilipi 1 year ago 0:55 for a fraction of a second I tought she would say the Protein is SuperImposed. uashuahsuhaushuah guess I'll let go of quantum mechanics videos for a while... If the double pendulum is "hard" to predict, protein folding must be a milion times harder. Thanks Sabine and the crew for always making such good videos. Reply @berniv7375 3 years ago Thank you for helping me understand a scientific problem that I never knew existed. Intriguing. Reply @Codethier 3 years ago (edited) your channel is a rare diamond Reply @JuBerryLive 3 years ago Very clear explanation. Thank you. Reply @ozhanfenerci7042 3 years ago Great explanations! Sometimes you may advise books and articles for further study. Reply @bobbymah2682 3 years ago Thanks for explaining this clearly Reply @Mrbillmacd 3 years ago Fascinating programme, thanks. Reply @davizitopa7252 3 years ago (edited) How do you calculate the incremental certainty of something? Like, let's say there's a set of elements (U, its count is denoted by C). Each of those elements has values (V1, V2, V3, ..., Vn) associated with a set of qualities all of them share (Q = {Q1, Q2, Q3, ..., Qn}). By knowing the value Va of the quality Qa, you know the element belongs to a specific proper subset of U: that of all elements which have value Va for property Qa, the count of elements in that subset is Ca. As you add the values of more properties to the same element, you get the intersections between the proper subsets, and the count gets smaller and smaller (like, let Sxy be the proper subset of U of elements that have value Vx for quality Qx AND value Vy for quality Qy such that Sx and Sy are both proper subsets of U and Sx and Sy are not equal). How do you calculate the count of subset Sn given that you know the values Qa = Va, Qb = Vb, ..., Qn = Vn? How to write that mathematically, and concisely is what I mean. Or, how certain can you be about an element En ∈ U, given that you know En has the quality values Qa = Va, Qb = Vb, ..., Qn = Vn? Reply @juanmarcos1145 3 years ago This woman is very intelligent, and she doesn't care about trolls Reply @AschKris 3 years ago I used to think that protein folding was more like folding sheets until I saw the great visualization in this video Reply @HatiKeseorangan 3 years ago (edited) Mem, i love how u present this Knowledge, my finger so fast to press subscribe button... Thank you... U make my research more better... I will watch another topic to, your topic is all very interesting... Reply @JuliusUnique 2 years ago good video! exactly what I was looking for! Reply @joewebster903 3 years ago Sabine the issue of fold or not to fold is resolved with supramolecular recognition at least for large protein chain used in polymers. Emil fisher discovered this biochemical relationship in 1914 and we use it today to solve issues globally with regard to enhancements to extend fortification and lifetimes of many condensation polymers. Therefore control of this process via the lock and key mechanisms for smaller proteins is equally possible Reply @NeilRieck 3 years ago I miss your public lectures at the Perimeter Institute in Waterloo, Ontario, Canada. Reply @KeithZSD 3 years ago A very sound summary. Although it's a very promising advent in solving protein folding problem, like ab initio in molecular dynamics, it will be still a long way to become a matured method to decisively determine protein structures. I envision for a long time this may stay as accessory tool aiding the structure determination from other means. Reply @googledev566 3 years ago Thanks for having subtitles... Reply @238assante 3 years ago approaching the 200k subs , congrats. I remember when you were at 25k , which is about when i subscribed. Maybe one day you'll beat Jeffree Stars, but you might have to start talking about makeup , explaining their protein folds. :D Reply @johnstonewall917 3 years ago Thank you for giving clarification to the claims of protein folding solutions. Reply @jjcm3135 3 years ago Truly amazing what random processes & mutations can come up with. Reply @alihouadef5539 3 years ago 3:28 Gold 1 Reply @Sirandar99 3 years ago A very interesting talk coming from someone who's used those tools in their simpler forms... One aspect you really should clarify is that understanding how proteins fold in an accurate and unbiased manner based on amino acid sequence is only a building block. Understanding how proteins fold and specifically designing an enzyme for an intended function are worlds apart.... Especially in the case you describe regarding enzymes that can catalyze the destruction of plastics. There is a myriad of microorganisms that stand to gain incredible evolutionary advantage if they could utilize plastic as an energy source or even a nutrient source.... If designing a plastic eating enzyme didn't require extreme thinking outside the box our plastics would already be degraded by microorganisms. In the world there is comparable amount of microorganisms to degrees of freedom of protein folding. Reply 1 reply @imaginographia2073 3 years ago Nice vídeos about new scientific discoverS! Tks, it seems Visualization in small scales (nano) plays an very important role in this era of science as well filters for big data(AI). Reply @yeshuabriles7862 2 years ago Thank you for sharing this Reply @gilbertengler9064 3 years ago Many thanks dear Sabine! Why should a protein after folding be at its lowest energy state? Indeed, a protein is made amino acid after amino acid and is probably continuously shaped based on a lowest energy regime. This however might impare to reach the correct shape at the end of its synthesis. Also its known that proteins, when gradually folding during synthesis, can interact with chaperon proteins to force a particular shape for the construction of the best "final configuration" to reach optimal activity. Reply @musicraftmenship4785 3 years ago This is an exceptional channel! Reply @DerTaran 3 years ago To be honest, when you announced, that you will get professional help with the look of your videos, I was sceptical, but wow, you can see the difference. 2 Reply @janerussell3472 3 years ago I might get back to folding-- and no doubt the Golden Mean will be involved-- but first a word on sampling size. The Chi-squared test is a popular approach when the sample sizes are large. When the large sample assumption does not hold, however, we need an exact testing method, such as Fisher's test. In geology, we find 10 samples are enough. With Ronald Fisher's test, the significance of the deviation from a null hypothesis (e.g., P-value) can be calculated exactly, rather than relying on an approximation that becomes exact in the limit as the sample size grows to infinity, as with many statistical tests. The beauty of Fisher's test is that it can be tested empirically against many examples from the same rock, not just 10. So we know it's reliable. Reply @stompcity4085 3 years ago Brilliant as always...I was glued to the screen. Xo Reply @luisfernandocuestasanchez4343 3 years ago Blessing young lady, thank you for sharing Reply @Alien0P 3 years ago (edited) I am an expert for "Bottoms up" approaches, especially with beer. Which also contains proteins Reply @jeffreyluciana8711 2 years ago I declare the science settled. There is no further debate Reply @maximusflightymus3892 3 years ago A Protein pallet should be designed to control initial folding, then extended to more complex control, Ai could be used to extrapolate the value of initial forming on the protein, to evaluate if the path could be controlled by early recognition of the pattern of development, maybe. Reply @wafikiri_ 3 years ago (edited) I have been using the internet since 1981, but when I found there was a world-wide web, and first used a navigator, the very first page I found (and I cannot recall how) was one of protein-folding simulations called Folderol. Their screen saver let my computer participate in the simulation. Reply @4thorder 3 years ago Very well presented and of course extremely interesting thanks! Reply @emiliocanals4520 3 years ago Her content is so high quality and makes very complex ideas accesible to anyone! Reply @TheLegionofReason 3 years ago NMR spectroscopy is also used for 3D protein structure determination with the advantage that it can be done for proteins in solution replicating physiological environments. Reply @DanieleMarchei 3 years ago Thank you for the informative video Sabine :) I'd like to dive deeper into this topic and RNA folding in general, could you list the publications you read during the making of this video? Reply 1 reply @zhiqiandu3110 3 years ago (edited) 3:32 Illuminating! Reply @kazimgumus7232 3 years ago İt was very informative. Thanks Sabine! Reply @mgmonteiro1 3 years ago I know some colleagues who are pretty grumpy about machine learning and sound quite funny when ranting. However, I don't think grumpiness is all there is to their concerns. First off, black boxes like Alphafold can actually "create" their own datasets using categorizing functions that determine the best folds among an initial set, creating the next best set for a second step, and so on. Only the initial conditions really have to be made from actual data or, if you're very nostalgic about statistical physics, just turn the entire thing into an elaborate Monte Carlo and shoot in some random folds :P People give fancy names to things all the time, with machine learning being the next rebranding for 'automated optimization', where the very parameters of the model can be changed on demand by the algorithm itself. In this case, this is a very "safe" application for machine learning, since if we choose bad categorizing algorithms we will find out (hopefully?) way before we inject proteins into humans, that the pretended miracle drug is actually a toxin. However, the same principles can be applied to all manner of things and already is (driving, surgeries, and soon even policymaking and legal inquiries), and I have to agree with my grumpy colleagues that we're not as good as the machine learning algorithms to fix their linear algebra mistakes on the fly, nor did we think of enough safety triggers when the self-driving car accidentally optimizes your lifespan. All things considered, it's inevitable that we are going to be using ML in more and more applications such as drug development/protein folding since it's quite simply the easiest solution to such complex problems, until or unless we find the nice equations that describe dynamics at this uncomfortable "not too quantum, not too classical" level while simultaneously being something we can reliably use on computers (and we have failed miserably so far). But fortunately to my grumpy friends, we can't really start relying on it alone anytime soon as well. Reply 1 reply @JoseAlvarad097 3 years ago going to school for this in a couple years! super excited :) Reply @jlpsinde 3 years ago Great as always, please keep your work! Reply @rockcrg3492 1 year ago This lady is wonderful. She is keeping up with relevant science. I love her intelligence and her digestible explanations!🫠 Reply @Osmiculture 1 year ago (edited) Protein engineering would've been my chosen field for a doctorate (but life got in the way). As an undergrad, I remember doing a dissertation way in back in 1990 for a GM course, naively looking at the initial technical framework and conditions required to create what I dubbed 'Plasticases'. I was interested to see that in the last decade, two separate cases bacteria have been found in different dump sites (one in Japan) that seem to have an effect on degrading some plastics, but we've a long and probably insurmountable journey ahead with all the different polymers that have been dreamed up by man. Reply @markangles3877 3 years ago Thank you Sabine! Reply @bocckoka 3 years ago Thanks for presenting, see you next week! Reply @Alondro77 3 years ago My lab is poking around with a mysterious mutant transcription factor that has severely diminished function, which I suspect is due either the mutant residue being either super-crucial for binding to partner transcription factors in the transcriptional activation complex, or it's the point of a crucial fold that is disrupted. A strictly conserved 'magic E' (glutamate residue) in a key sequence motif, when mutated to ANY other amino acid... drops the activity of the transcription factor to almost NOTHING. We haven't tried aspartate yet, however. Aspartate would be predicted to have some activity (at least 50%) since it has the same functional group but a one-carbon shorter side chain. This might NOT be the case if some type of unknown enzymatic activity is taking place there, and that may be the case, since I've noted that in the predicted 3D structure, there is a histidine and a serine in close enough position that it may make a Glu-His-Ser catalytic triad. This would be quite a discovery for a transcription factor to also be a functional enzyme. Reply @SzTz100 3 years ago You cover every interesting topic that exists. How do you do it and where do you find them? Reply @flugschulerfluglehrer7139 2 years ago One year later the accuracy of alphaFold is at 90%. 1 Reply @animavideography1379 3 years ago (edited) When I completed my BSc in Biochemistry in 1986 Sabine I could not in my wildest dreams have even imagined that AI would ultimately have been the key to unlocking this seemingly unsolvable mystery. Reply @leodiamondlegacy 3 years ago Hello I’m new to your channel, but have been thirstily searching for content that addresses difficult topics without “dumbing it down” too far or without assuming the best place to start is the beginning. The beginning is not always the best place to start and I appreciate that you waste no time in getting deeper into the subject. For example. If i click on a video addressing the question of wether or not time is real, i can very easily do without the first typical leg of the journey; were we discuss how long in history philosophy has wondered this, how the 19th and 20th century brought this into new light, and i can certainly go without the whole “have you ever really thought about it?” Moment... I don’t need to hear for the billionth time how einstien dethroned newton. I think that educational content on youtube should be categorized into beginner, advanced, and so on, and i think your videos seem to fall nicely just beyond the beginning.. You do not talk down to me, you assume i have an ability to “keep up” and I appreciate this greatly. I will be commenting very very little but i want you to know that i am watching everything you upload with interest. Thankyou. Reply @Ireniicus 3 years ago wow, what an interesting subject that could positively impact everyone and many other lifeforms in the medium term. 1 Reply @richdog490 3 years ago It should also be noted that only something about 1% of all proteins are crystallizable. Most proteins don't crystalize, and the structure of the protein whilst inside of a crystal structure comprises the vast majority of the data that has been used to train the neural network. Reply @benmcreynolds8581 2 years ago When i learned about prions and just the utter amount and possibility of proteins (the usually helpful things) can turn against you, and you can get improper protein issues and then suffer from a number of health disorders... I was like nearly speechless and can understand why someone might want to live in a bubble. Like bubble boy lol Reply @janerussell3472 3 years ago HIGH CHAPERON ITERATION, BLUE BOY? Larger proteins, especially those with complex native state topology, are often kinetically trapped in metastable states for sufficiently long times that protein aggregation could be a major problem. That's where chaperones come in. Chaperones drive their substrates out of equilibrium, and in the process maximize the rate of native substrate production rather than the absolute yield or the folding rate. For example, the GroEL and RNA chaperones involved in unfolding of the substrates suggests that there ought to be a universal mechanism of chaperon-assisted protein and RNA folding. Unlike the cage model, the Iterative Annealing Mechanism (IAM) quantitatively explains all of the available experimental data, including the effect of GroEL mutants in the rescue of mitochodrial Malate Dehydrogenase (mtMDH) and Citrate Synthase. Blue: I mean no disrespect, ma'am, but you could never be my mother. Victoria: Never is like always, Blue. There is no such thing. Reply @alecouto 1 year ago (edited) Great vid Sabine, do we have any recent development on this matter? Reply @Lucerozeus 3 years ago I took biochem in my university and it was my favorite class of all time. Great video! So interesting. 1 Reply @alihouadef5539 3 years ago 3:28 Gold 1 Reply @KeithCooper-Albuquerque 3 years ago Thanks Sabine for another great video. Thanks for saying your name so often in this one! BTW, nice outfit once again! 1 Reply @Shaffaqwamiq 3 years ago Great explanation. I had no idea what this was. Subbed to the channel too. :) Reply @mmicoski 3 years ago The numbers of possible ways proteins fold is big even comparing with the Universe age. If one fold is tested each second then we can define 1U = 1 universe age in seconds. In this case, 10^150 ~ U^9 and 10^50 ~ U^3 Reply @pericles2122 2 years ago many thnx, Sabine. Reply @maximpavlov5379 1 year ago "That's almost as many vacua as there are in String Theory!" - this cracked me up! :) Sabine and her undying love for 10^500 universes in String Theory! :))) Reply @badgerlife9541 3 years ago Your videos are very informative! Thanks! New subscriber here Reply @StewartChaimson 3 years ago You're doing a great service! Reply @subathrann 1 year ago thanks for the information Reply @nynra6584 3 years ago (edited) Thanks to creators like you I can impress my friends with the knowledge that I learn Reply @nitesan2814 3 years ago Great vid as always ma'am Reply @davidblanton5407 3 years ago thank you for your presentation. It is lucid and enjoyable. Reply @somewherenorthofstarbase7056 2 years ago "That is almost as much vacua as in string theory!" I caught that! 1 Reply @ashwinkolhatkar9126 3 years ago Awesome explanation thank you! Reply @Bareego 3 years ago (edited) As someone who has dabbled in protein folding at foldit, I suspect that this AI approach will surely help a lot in finding protein shapes that depend on other proteins folding a similar way. But with your secondary and tertiary structure you'll sometimes get big surprises how it folds completely differently than you might expect. That said, I really hope that this AI or similar approach will eventually lead to a predictable and predictive way of getting to know protein structures from the DNA that we know. I don't think many people appreciate what a violent wobbling place a cell is. Things bump into each other all the time at high speed. Also the possible structure numbers posted might be a bit blown out IMO. The dna sequence and resulting protein chain puts somewhat of a limit on angles and tension in the resulting protein, at least in many positions of the structure. These are not obvious to many simulations though and might gain a lot from having an AI approach towards getting a protein that has a low backbone tension and gets close to the shape of the protein in reality. Reply @AnsImran 2 years ago lolzzz! 3:32 I was also thinking the same. 'That's almost as many vacua as there are in string theory' Reply @anonymous.youtuber 3 years ago I’m so glad I found you ! Reply @mercster 3 years ago Back in the day (after SETI@Home wasn't interesting anymore) we used to run Folding@Home. This is interesting! Reply @BBQDad463 3 years ago Thank you for this excellent video! Reply @nsa8933 11 months ago one is the protein. A.I, is the product of a short length protein. helping the problem ( which is A.I) not the protein itself, (which is split in 3, from being X.1) Reply @lucasfc4587 3 years ago The explanation was phenomenal, with only some advanced highschool knowledge, I could understand 99% of what you were talking about. Now I’m smarter, thanks to you! I can’t really make us even, but can at least express my gratitude :) Reply @danboruman5039 3 years ago Thank you for the piece. Just a question, is there a rigorous way to analyze the forces interacting in certain domains of proteins? Reply @williamschacht 3 years ago It's nice that you talk about Stephen in the present tense. I miss that guy. As excited as I am about scientific computing, mathematicians would say that it's not even in the ball park. 1 Reply 4 replies @tsawy6 3 years ago Once we get past the problem of figuring out what a protein looks like, the next logical step seems to be figuring out what it will /do/. It seems like this would be essentially required to actively engineer many (but not all) protein effects, especially enzymatic ones. Assuming Deep Mind, with a little more work and tinkering over the next, say, decade, could get nearer to, say 99.9%, how much more work would we have to do to get here? Reply @iwanabana 3 years ago Fantastic stuff, Sabine. Oder wie man sagt: krass!!! Reply @benotyourboss 3 years ago Nice video Sabine! this is really remarkable, I mean it would take a 100 amino acid long protein with 3 degrees of freedom and the ability to "check" 10^12 conformations per second more time to find the right conformation by random approaching then the universe has seen, however such amino acid chains find there conformation in seconds :-)! nature is astonishing and that is one reason for me to study sciences! Reply @johnd9031 3 years ago Science is amazing in the scope of problems it can tackle. Reply @nia6849 3 years ago Proteins are complex molecules determined by the sequences of a.m (amino acids arrangements). 1 Reply @gwiyomikim5988 3 years ago Maybe it’s the times in which we live but my first thought upon hearing of DeepMind’s recent success in the CASP competition was did they made a breakthrough, or did DeepMind’s owner, Google, merely find a clever way to cheat. Given Googles intricately folded tentacles into every online data platform in the world would it be that surprising? 2 Reply 2 replies @pauloemanueldeoliveirafrei654 3 years ago Congratulations from Brazil Profa. Sabine.....I would like to hear you explain the oldest discussion in Physics ...LONGITUDINAL AMPÉRE FORCE ...( Peter Graneau, André Koch-Assis, et al ) ....I have made some experiments with L.A.F. and really it is difficult find a correct explain in nowadays Electrodynamics ( Maxwell/ Lorentz ) ...thank you ! Reply @vincentconticello2613 3 years ago Hmmm...interesting topic. I have heard of the results but have not studied them in detail. The larger question is whether the approach can predict the structure of folds that have never been previously observed, unfortunately there is a diminishing number of those. It also finesses to some degree the question of whether in vitro determined structures truly resemble the in vivo structure, which remains an open question. However, when a scientist of the stature of Andrei Lupas is impressed, reasonable people should pay attention. Even if the program could “only” predict the structures of known folds from sequence data alone, it would be incredibly valuable. If it could move forward into protein design and the prediction of unknown structures, the so-called dark matter of the protein universe, then it would be truly transformative. Reply @Old299dfk 3 years ago 0:10 - wait a minute... the 'KRAUT' science app?! Haha! God I love Sabine. Reply @parrotraiser6541 3 years ago Media love to hype "breakthroughs" and panaceas, (which usually disappoint), but it's good to see steady progress being made. As one of the other comments points out, taking care of the routine cases lets researchers concetrate on the really gnarly remainder. Reply @AnexoRialto 3 years ago I had only heard a quick hot take about the success of AI in protein folding. Thanks for providing a much more complete and understandable explanation of the topic. Reply @n16r3d0 3 years ago (edited) CASP in a nutshell: 1. Let "teams" of natural intelligences (humans) form randomly (via unpredictable historical processes). 2. Let each team produce a model of artificial intelligence to predict protein folding (using open data sets). 3. Test each team's model against a "fresh" validation data set. 4. Reward the teams depending on their models' performance. 5. Rinse and repeat in two years. I just freaking love how meta this is. Reply @volleyballschlaeger 3 years ago I can't explain why but the NordVPN commercial at the end looks like comedy. The commercial is the part of this video where you are not playing your own role but some role for NordVPN. Reply @Football__Junkie 3 years ago I went to Home Depot the other day, bought some disulfide bonds, and did some protein folding Reply @brendakrieger7000 1 year ago Thanks for the explanation! I have Type 2 Diabetes. Reply @vperez4796 2 years ago (edited) As a protein is assembled, there are "chaperones" that help to cast the "right" folding. It is also known that chaperones hold tightly a misfolded protein and destroy them. There are states of the same protein, one is the native-protein the others are inactive foldings. Chaperones know which conformation is right. Shoudn't they study chaperones as well? Reply @bloodworthmagic 3 years ago Sabina reminds me of the superhero costume designer and manufacturer from The Incredibles movie Reply @rahulsood3412 2 years ago Thank you for an insightful video. While protein folding problem is being resolved, I believe that it is still early days in study of proteins (& their changes). Proteinomics has still not reached the penetration which gene sequencing has. In light of the same, would you agree that companies like Quantum-Si (Founded by Jonathan M. Rothberg) are making a huge difference to first understand proteins by detecting the underlying amino chains ? Reply @justinpatterson5291 3 years ago You've just reminded me of what I can use my excess gpu cycles for, while I sleep. Reply @edreusser4741 2 years ago The folding can also be affected and changed by altering the Ph of the solution it's in. Reply @aresaurelian 3 years ago Neural Network and proper applied artificial intelligence can be used to predict the entire universe. And it can also be used to predict and design measuring devices that can collect all this data it needs. 1 Reply @charleshendry5978 1 year ago Excellent, no mention of "Folding at Home"? A distributed program and its effects if any? Reply @mbdtsmo 3 years ago Loved that little stab at string theory :) 1 Reply @ChrisM541 2 years ago Current AI is all about pattern matching - take an unknown and see if it can be matched against a database of 'known'. Unfortunately, therein lies it's Achilles Heel. If no match is found (mainly because the database is too small) then what do we do? Either we say "match = 0", or, we introduce guesswork by returning a match percentage score e.g. 50% = half of the amino acid (AA) sequence has a match with one/more known sequences...but, there is a problem here - do we mean half the 'linear sequence' is a match? (fewer returns, possibly higher quality), or do we mean multiple smaller sequences(>1 AA) have matches, and when combined, equate to half the full sequence? (more returns, possibly poorer quality), but makes the algorithm look better than it is(!!). Add into this mix is the fact that there will likely be single/more lengths of the AA chain that can be substituted with other specific AA's and not have a detrimental affect on the folded protein's full functionality. I'd like to think that lots of work has already been done on the effects of AA substitution for known proteins. Then again, it is indeed, a hugely complex subject, one we have barely started to understand. I really wouldn't put too much faith on current AI. Reply @michaelg9344 3 years ago My can of spaghetti and meatballs froze in my car, thawed and seeming coagulation gave a different texture to the food, most unpalatable. Long chains of protein lines are reduced space into globular rounding, reducing the long sequence into a folded twist of the same molecular order. Reply @merbst 3 years ago (edited) Neural networks are approximation algorithms! An NP-Complete Combinatorial Optimization problem has no polynomial time solution unless NP is proven to equal P. Reply @pokkiebruno242 3 years ago Nice String theory burn 3:30 Reply @rollmeister 3 years ago What about Kuru or Creutzfeldt–Jakob Prions? No mention of either. Protein folding depends on complex chemistry bonding & instance specific. Reply @jamesruscheinski8602 3 years ago Keep updates on protein folding coming Reply @hmgrraarrpffrzz9763 3 years ago Thank you for the video. I'm normally not big on subscribing to channels, but you get to be the third channel to be added to my subscription list. :) Reply @nijimbagao7090 3 years ago Watch for the topics stayed for the science... congrats Sabine I just suscribed and will get you at least 5 more suscribers... that is exponential but only on a cooking program way Reply @uhmnope4787 3 years ago I hear NordVPN makes generous offers. Hope they gave you all they can, you deserve it! Reply @uoabigaillevey 3 years ago While I am not a scientist in any sense of the word other than just being interested and minimally knowledgeable beyond Biology 102 I do find this not only very interesting but also very heartening for the future. Given that v2 of the software included improvements to raise accuracy, just imagine what improvements the eventual v3, v4, etc. will bring. This will greatly speed up development of countless biological applications and will mean lives either saved or improved through this technology. I am rooting for those involved to have much future success.. for as they succeed we all benefit in the end. Reply @steveschafer5626 3 years ago Loved the string theory jab! 1 Reply @sweiland75 3 years ago I once had the Folding@Home distributed computing program on my computer. I don't even know if reinstalling it now would be worth the work. Reply @JimNichols 3 years ago Please stop jumping from side to side to middle of the screen, it is most difficult to watch. I love the videos ma'am, I realize they are difficult to video and edit but the subject matter you choose in these is most interesting thank you :) Reply @user-cl5vj8nu7c 3 years ago the best sober science channel ever! Reply @thebeelight 3 years ago this is better than DeepMind's own version! Reply @DrakiniteOfficial 3 years ago Don't have anything to comment on, except that I really like your videos, so here's a comment to boost engagement Reply @pablomaster4225 3 years ago (edited) Very well explained subject !! It was as professional as it can get and its on Youtube for free!! Unbelievable! Id support you on patreon if I wasnt so broke nowadays ;C Liked and Subscribed - its the least i could do ! Reply @talleyhoe846 3 years ago Another educational treat! Reply @yoramalon5273 3 years ago Inspiring video. It is just a matter of time until A.I would dwarf the human mind. Every time that a software, has ALL the data, it beats the human mind. Chess, stractural optimization, smelling, hearing, path optimization, analytical chimestry, face recognition, finger print recognition, reading... The human mind is finite and has its own limitations. Reply @kori228 3 years ago while I'm not an expert by any means, I've been following some videos on Type 2 Diabetes, and I wouldn't say it's caused by improperly folded proteins, but rather caused by hyperinsulinemia and the body shutting off insulin receptors as a countermeasure to the high insulin Reply @bscher5003 3 years ago Thank you Sabine. I believe semiconductor Quantum Dots (QD) emitting various colors can be attached to protein and DNA strings at specific genes, such that coincidence of the detected colors are indicative of the protein folding sequence. Using spectroscopy instruments, defective genes can be determined by detecting the overlap coincidence of the colored QDs and their folded positions on the sequence. Reply @billyt8868 3 years ago (edited) what’s most interesting is that under standard biological conditions (temp, pH, osmolarity, sometimes voltage) there is 1 single proper folding configuration. it’s nearly impossible to predict the structure but they very reliably fold nearly perfectly. Reply @admiralhyperspace0015 3 years ago Awesome video. Got my sub. Reply @Faustobellissimo 3 years ago Hi, Sabine, another delightful video! A question about biology: have you ever heard about Rupert Sheldrake's hypothesis of "morphic resonance"? If yes, what do you think of it? Or would you make a video about it? 2 Reply @ruschein 3 years ago @Sabine: Hi, do you think that you could make a video about hidden variables in quantum mechanics and Bell's theorem? Reply @john_hind 3 years ago I can't help thinking that the CASP question "Given the amino-acid sequence, compute the most stable protein shape" is the wrong question. Nature can already do this and it constrains it to the random search and directed selection toolbox of evolution. If you want to do bio-engineering by intelligent design, surely you need the reverse computation "Given a required protein shape, compute the most efficient amino-acid sequence that will fold to that shape". Is the AlphaFold 2 process likely to be reversible and therefore give us a toolkit not available in nature, or is its importance limited to simulating natural processes more quickly or inexpensively? Reply @LoraxChannel 1 year ago Good summary, but why omit Lvinthal's paradox and it's implications. Reply @fidelogos7098 3 years ago As someone with a 40-year-old degree (unused) in biology, I was happy to find this channel. I've never lost my interest in biology, specifically molecular biology. If protein folding has indeed been solved, the ramifications are endless. Could there come a time when disabilities and diseases are a thing of the past? Reply 1 reply @dr.jamesolack8504 3 years ago Nice jab at string theory.....please, keep stirring that stick in the ant bed!! I ️ it, doc! And ️ your channel.! Reply @LabyrinthMike 3 years ago Well, that was extremely interesting. I have heard about protein folding, but didn't understand what it was before. Thanks! I must admit that I'm surprised that, if one knows the sequence of amino acids, that one should be able to write a program to mathematically construct the folded shape. It sounds pretty deterministic. But apparently, this is not as easy as I think it would be. Reply 1 reply @chrisanderson4516 3 years ago (edited) The protein folding problem has revealed much about what happens when proteins do not fold properly. It is detrimental to the cell. It creates different diseases, etc. Proper folding indicates the proteins are designed. For when mutations occur (improper folding) aka evolution, then detrimental effects occur within the cell. Only with a proper folding of proteins will cells function to their highest potential. This indicates proteins are designed, not from random mutations. Reply @timmyjones1921 3 years ago I've subscribed now , I think Sabine could really have a go at Sabine's Fashion Line to go with the wonderful music she does also . Now On To Protein Folding. Reply @kkkiren599 3 years ago Humans got to assess the adaptive foldings of proteins related to the " functions" they perform. Their contact " sensors, carriers, and other allied parts assume the spatial ergonomic process they have to perform. They alter foldings to suit their functions. The amino acid sequence and folding may correlate to their precursor origins or progenitors based on threshold responses. Nothing functional is wasted in " efficient" nature although redundancy adaptations may arise ( as in immunology). God bless Reply @carloc352 3 years ago Awesome video. This tells me that physicists are not only the best engineers, but also the best biologists (I’m an engineer and I worked with a couple of physicists). Of course Sabine is a top physicists. Reply @guiasparaservir8792 3 years ago I loved this video!!! Thanks! Reply @happyhome41 3 years ago "Surprise" topic - excellent scratch of the surface. I feel smarter today. Reply @brothermine2292 3 years ago What about the other half of the protein-folding problem? To be able to find a molecular sequence that would fold to a desired structure. Reply @oloidhexasphericon5349 3 years ago (edited) 5:25, this feels like a google interview question Reply @Earwaxfire909 3 years ago I did some of this work back in the early 1990s. We predicted the structure of a new interferon compared to a known structure. We got a result that was useful in understanding what mutations were doing to its activity. And it was obvious that this process could be advanced with more data and greater computing power. I do feel that this new approach is great and powerful. It will be very useful in understanding critical regions of proteins. And it will begin to explain some aspects of protein function, Hurray! But the truth is that it is more like Kepler than Newton. And oh boy do we have a long way to go. But yes this is great work. Reply @vperez4796 2 years ago (edited) The AI models are predictions based on alpha-HELIX and beta-SHEETS. A concept they feed to AI during programming AI. AI is based on known existing x-ray diffraction patterns. Those "AI predictions" are not only based on SOLID state X-rays but also they are structures IN VACUUM. The AI predictions also lack of "water molecules" and polarity of the surrounding. The alpha-Helix together with beta-Sheet folding are a primordial templates that help fold the proteins in a way that doesn't occur in the vacuum. Furthermore, 200 molecules of water or more are NOT contemplated by AI models. H2O interact with surrounding amino-acids "residues". There it is a 4th dimension of the protein folding, the CHAPERONES. This complicates matters further, cast of other proteins prevents unfolding. It was AE who said "Everything should be made as simple as possible, but not simpler". AI could not come out with that from the vacuum. Just a computer hype of AI. Take a look at this presentation: https://www.youtube.com/watch?v=1peFJ_-N7V8 Reply @value8035 3 years ago Vacuas in string theory.. ? Ah.. that much.. When Dr.Sabine draws metaphors... Reply @SpectatorAlius 3 years ago Sabine is quite right, of course, to point out DeepLearning's deep dependency on a data set. But there is an even more fundamental limitation to most, if not all A.I. When it gets the right answer, not even the people who designed the neural net or tuned the algorithm have any idea why the A.I. device got the right answer. In fact, this is one of the danger's of the movement to A.I. We don't really know why it works (when it works), so we cannot explain what went wrong when it fails. Worse yet, since programs have no legal liability, and generally, neither do the developers, this is going to produce an avalanche of accidents in which nobody can be sued, so there is no financial motivation for considering the safety of the users and public paramount. Reply @expertnonexpert885 3 years ago could you do a video on the magnetic field and how it seems to do work but doesn't? this is the most difficult thing for me to understand in physics so far. also the H field. Reply @homejonny9326 3 years ago Thanks for the amazing video. I would like to add that not necessarily the AI needs data for the "new cat protein". From what I've learnt by popular science sources, the huge amount of data handled by AI sometimes has the feature of predicting results in "new areas" as well, such is the power of the AI eyes. Reply @TheFarmanimalfriend 3 years ago My senior research for my degree in biochemistry was about protein folding. Chaperone proteins are essential to direct the protein folding as it is extruded from the ribosome (where the t-rna is translated into protein). She doesn't mention chaperon proteins which is unfortunate as they are key to directing the protein to fold into biologically active form (enzymes). procaryotes (e-coli) do not use chaperone proteins and cannot be used for creating biologically active eukaryotic proteins. This fact saved the human race from a recombinant nightmare. 1 Reply @bosoerjadi2838 3 years ago 10^300 being 'almost' the number of plot holes in string theory.. That's not just harsh. It's a burning fury. Reply @Mareczekw30 3 years ago Next great challenge will be virtual cell. Problem is, it is estimated to be completed in 50-100 years from now. We need beter biochemistry understanding. If we will be able to finish this, lab tests will be just formality. Reply @srobertweiser 3 years ago When we were little, me and my brothers and sisters had a nanny from Guatemala and she used to fold all of our proteins with our laundry. Reply @sanesanyo 3 years ago Danke Sabine. Viel gelernt. VG aus Frankfurt Reply @gewinnste 2 years ago It's actually pronounced pro- teen (like in e.g. teenager). Being German I made this mistake myself until a native speaker made me aware of the right pronunciation. Reply @nziom 3 years ago (edited) this is promising considering the current situation Reply @lel6360 3 years ago I was surprised that you were not more critical of the claim that protein folding has been "solved". While I think deepmind achieved something amazing, as far I understand, there is still work to do. There is the missing 10% as mentioned in the video. Many proteins fold into shapes that interact with molecules in dynamic ways. Helicases spin at 10,000 rpm and DNA polymerase moves strands of DNA around to help with the replication process. It's not clear how much useful information AlphaFold II could give about these kinds of proteins. Sometimes proteins are modified after folding, for example insulin, where a strand folds, bonds with itself, and then some parts are removed. The AlphaFold II paper is not out yet, but it is almost certain that it cannot properly handle proteins which lose some amino acids, because exactly which parts are cleaved will depend on the other proteins in the environment. AlphaFold II uses information about the evolved changes to proteins. It has also only ever been given a biologically functional protein as training data. This means it is unlikely to generalise outside the domain of biologically functional proteins. This is of course a huge and important domain, but it means that truly novel proteins would not be reliably folded by AlphaFold II. If AlphaFold II could predict the 3D shape of (more or less) any amino acid sequence, it could search for ones with binding pockets that fit transition states for chemical reactions that are useful to us, for example those used in industry, or the degradation of plastics. AlphaFold II can (probably) only handle the 21 biologically common amino acids. If protein folding were truly "solved", I would expect that more or less any possible amino acid would be acceptable. My point is that while this is an amazing achievement, I think the claim that protein folding has been "solved" is highly misleading. Can AlphaFold II correctly predict the structure of alternating phenylalanine and cysteine residues? (Something that is likely very different from anything in its training data). Could AlphaFold II, paired with anything else available today, design a novel DNA polymerase? I would call a problem "solved" when there was no longer anything significant to find out about it. 1 Reply @Deciheximal 2 years ago "Prions" for cow sickness was such a bad name choice - takes up a name that would much better be used for particle physics. Reply @tezlashock 3 years ago (edited) Can we model this behavior by analyzing the fractal dimension of the 1d chain? Also I think that proteins fold in one specific way in one specific environment and there are many, many, many possible 3D shapes of proteins Reply @whatwhat9519 1 year ago (edited) Prion diseases and rabies are the only two things im afraid of Well besides being afraid of everything else as well 1 Reply @Inpreesme 2 years ago Thank you Reply @bradsmith9189 3 years ago No. Information is by far the biggest problem. Where did the information for the first living cell come from? Where do the information to build new animals during the Cambrian explosion come from? Etc etc... 1 Reply @pauldow1648 1 year ago Thank you. Reply @linkin543210 3 years ago I don’t think the 90% figure applies to new protein sequences it has never seen before. In other words, it was simply modelling existing parameters and finding a best fit model but not necessarily learning/understanding how it happens. Reply @ObviouslyCrap 3 years ago This explains why my omelettes always resemble scrambled eggs, i.e., improperly folded proteins. Reply @KipIngram 2 years ago But even if AlphaGo reaches a reliable 100%, will that really mean that we "understand protein folding"? I'd say not - it's not necessarily possible to understand how deep learning algorithms perform their feat. What I consider the proper goal is a clear physical understand, from first principles, of the folding process so that we can arrive at the correct answer by a direct physical calculation. 1 Reply @shaileshrana7165 3 years ago Headline: Protein Folding has been solved. Me: Yes! I can finally learn how to play Sudoku better! 2 Reply @ascaniosobrero 3 years ago Interesting review paper on the physics protein modeling (and CASP) in Science Nov. 27 2020: Brini et al., Science 370, eaaz3041 (2020) 27 November 2020. Reply @muuubiee 3 years ago You don't actually need data, deep learning generates that itself... But you do need an environment that can produce results. So if we can calculate the environment ai will be able to come up with any proteins. Mufold is probably gonna improve even more, as it'll be less restricted by coded rules. But what I get out of it, is that we can't compute the environment. Reply @aniketchandra6492 3 years ago Remarkable video!! Reply @frogandspanner 3 years ago 10:26 AI may be able to predict a tertiary structure from a primary structure, but is that reversible? Given a required function one could postulate a tertiary structure to achieve that function, but could AI then reverse generate the necessary primary structure? 1 Reply @ramonmartinez1348 3 years ago We’ve got to learn more Reply @onehitpick9758 3 years ago I had some great protein folding software back in the late 1990s, and figured it was not a big problem. It's good to hear it's not. Reply @Sosukz 3 years ago Finally the algorithm is working. Reply @lalalafamille 3 years ago Great video thank you Reply @tomcan48 3 years ago GOOD TO HEAR THAT AI HELPED TO BEGIN TO SOLVE THE PROBLEM OF CONSCIOUSNESS FROM CELLS, BEING OBSERVED Reply @jjeherrera 3 years ago Regarding the fear of people who may feel they'll be rendered obsolete due to artificial intelligence, they should consider much of the boring work on data analysis they do can be left to AI, while they deal with the main issues. Reply @rickvassell8349 3 years ago I'm glad there are a lot smarter people out there than me. Reply @djehuti3 3 years ago superb, thank you Reply @alleneverhart4141 3 years ago (edited) Sabine, great video! Do you think that maybe AI can solve the quantum-gravity problem? 1 Reply 1 reply @dreamarcher4018 3 years ago It seems to me from what I recall about prions is that they are misfolded proteins have they figured out how to make them disappear or how to prevent that. Otherwise we will still have Creutzfeldt-Jakob disease in humans and chronic wasting disease in deer. Reply @andyl9900 3 years ago Would love your opinion on the AI singularity sabine 1 Reply @DaBlondDude 3 years ago If I follow properly, this seems like high end pattern recognition. If the shape affects the function and the sequence of amino acids limit the shape, is there only one possible solution for any specific sequence? Also, do we know how to sequence these amino acids in order to customize proteins? Reply @tmst2199 2 years ago I think you could make just about anything interesting, Sabine! Reply @davidderidder2667 3 years ago Excellent content Sabine. Can you please keep us updated on how AI techniques help us solve this problem. The funding of more science and better science education is also a very critical problem. Would love to hear your thoughts on that. I will also look into the NordVPN product. Reply @rdjinaz 2 years ago Folding proteins appears to be a "hideously difficult" task for nature, too, it would seem. Cells cannot always "expect" proteins to fold into their proper confirmation, based solely on the amino acid sequence they assemble at the ribosome. In some cases, the intracellular machinery of employs the assistance of "chaperone proteins" (and even co-chaperones) to ensure protein conformation is physiologic. I wonder if this doesn't mean that AI won't be able to predict the tertiary or quaternary folding patterns in all proteins, based solely on the amino acid sequences? Will we soon be using AI to design protein chaperones to coax proteins into "behaving"? Reply @Kosnoros 3 years ago (edited) A great topic for another video would be to explain why it is unfeasible to calculate protein folding from first principles. Why not just take the Schroedinger Equation put in the potential term for your chain of amino acids and use simple numeric integration to solve it? I was always wondering why that doesn't work.This might be part of general video on the limits of numeric computation. Thanks for the great videos. Reply 3 replies @sukalpashrestha61 3 years ago Didnt know character actress Margo Martindale was into protein folding Ps. Good video btw 1 Reply @2nostromo 1 year ago Many factors affect the folding process. For one thing it is a temporal process with secondary structures folding as the aa chain grows. The local pH and ionic strength, even the temperature and changing dielectric micro-enviroment comes in to play. And who says the most stable structure is the most biologically relevant? If so Wetlaufer's reagent would have been more successful. The AA sequence is only a starting point. Reply @bhangrafan4480 3 years ago One of the problems, as you mention, is whether the biologically relevant structure is globally stable or just metastable. The existence of prions suggests that some biologically relevant folds are only metastable. There are other reasons for thinking this as well. The main approach has been molecular dynamics simulations, but this uses a lot of computing power, as in the development of IBM Blue Gene. Perceptron training is an approach which has been tried for some time. Reply @PhysicsNative 2 years ago That 10%+ is key chemical physics, not programmed or soluble by the AI routines. XRD and other methods will be around for some time, indispensable. As a commenter below stated, reaction rates are determined by precision locations of small constituents. Reply @thelocalsage 3 years ago Great video ! 1 Reply @DannyAlvaro 3 years ago como siempre excelente video. Reply @omarabdelkadereldarir7458 3 years ago (edited) Great video! 1 Reply @sakubashiba3610 3 years ago i think i got this solved...so lets say that proteins fold but make that in time from 0 to 5 so it starts as a wave function for straight protein but it needs to change dirrection up and start again a straight line so the wave needs to be smaller or bigger for left or right turn...what im getting at is this... you can simplify the whole folding process into a simple 010020300050400010010101010100010101000 type of code... like if you draw it into a x and y where it goes from 0 then to 2 then to 0 then 2 its a wave that wave can be translated to the protein folding...also its possible when proggrammed correct to input like 7 and 8 for special fold left fold right type of thing... what im trying to say is that if you have a number lets say "1250001050500250042305020070001002400412040602060...." and on that number can be translated into the whole protein....the thing is this can actually be programmed and decoded... also it can be stored... because lets say you got a number for protein X12 or whatever that is 921 digits... using a decoder you decompress it you compile it and then you can even place it on a graphical interface... i think this if programmed right would be able to get all proteins available in the universe... or all the folds possible....untill lets say 10000 digits... remember digits then can be translated to chemicals if needed ... this gives the logic only....also if you take the folding patterns.. and somehow use it to manipulate this 10000 digit code you can get folding in future or past for any protein....wellp i guess its solved then... Reply @dr.michaellittle5611 3 years ago Superb video.️ Reply @Flash_345 3 years ago What about using tools like molecular dynamics to study protein folding? Reply @MitchCrane 3 years ago Turning cats into turtles? That's the real singularity. Reply @allenrhoades8482 3 years ago By what mechanism does natural selection create anything? Isn't it the means by which a more favorable existing quality continues? 4:24 "This is probably because natural selection preferentially made use of large molecules which reliably fold the same way." Reply @LinseyProgress 3 years ago Bucket list, dinner with Sabine Reply @n20games52 3 years ago I've been following DeepMind's Alpha Zero chess playing. It's amazing. Reply @JustMike2791 3 years ago A 12:00 minute video has reaffirmed my suspicions that I am a moron. Reply @GolfSmooth 3 years ago Thank you Sabine. Do you think "deep learning" will ever go beyond shape recognition? 1 Reply @rc5989 3 years ago (edited) More great content from Sabine! If only we could understand what AlphaFold2 was doing, then progress would be quick, imo. AlphaZero plays chess in a manner that is superhuman, but comprehensible by top chess masters, from what I have read. Reply @WilliamFord972 3 years ago I’m not keen on Google entering my field (biophysics). The reasons should be obvious. 1 Reply @ivanhardman4576 3 years ago Why predict the way proteins will fold ? Is it not easier to train/force the protein to fold into a desired shape ? Reply @janerussell3472 3 years ago Assuming there was no garden of Eden, when did we diverge from chimps? But why assume we diverged from chimps? They have 24 chromosomes. In 2013, it was shown that the alleged interstitial telomeric repeat site of the human chromosome 2 fusion corresponding to chimpanzee chromosomes 2A and 2B of a hypothetical common ancestor was actually a second promoter in the DDX11L2 long non-coding RNA gene. In 2020, Piero Fariselli et al., talked about 'DNA sequence symmetries from randomness: the origin of the Chargaff’s second parity rule'. I'm not a YEC- the geology shows we're not under10,000 years old; though the Lead Paradox implies we're not 4.6 byo. Zircons would be turned over 9 times in that period, and would not have escaped unmelted. From the work of Jean-Claude Perez I see only humans ( not even Neanderthal's ) are on the Golden Mean for the whole genome. If I triangulate the common Eve with the common Adam I get roughly 185,000 years. Even Dawkins admits an original Eve and an original Adam, though not necessarily together. lol. Natural Selection obviously occurs...that's a pruning process. Evolution should have been put to bed with the discovery of orphan genes in 1996 i.e. gene sequences unique to the species. That's all. Reply @michaelhaas4431 3 years ago I find this very interesting. Do you know how they take into account that chaperons could help folding proteins they right way just as they are synthesized? It could mean that the biologically active shape of a protein can not always be deduced by looking at the final amino acid sequence. Reply @jonpowell6770 2 years ago i solved the protein folding problem by attending a chef school and learning the proper way to cook a western omelet! Reply @radwizard 3 years ago Hi, can you please do a video deriving the Gobbledygook Operator from first principles? Asking for a friend. :) Reply @13e11even11 3 years ago I would agree the AI has made great advances in predicting folds, but the problems seems unsolved to be what we need. Reply @ElonizuMuskamoto 2 years ago Could supplementing EAA‘s help me keep my proteins clean? Or could it even raise the risk? Reply @Jonathan-rm6kt 3 years ago Question: how are researchers encoding "shape" for these complex, non regular forms? Is it an actual geometry? Or a deterministic procedure? Reply @lollsazz 3 years ago But... What about post-translational modification? A lot of polypeptides get modified with sugars, which affect their folding Reply @redx11x 1 year ago How does DNA instruct a cell to make specific proteins that are folded in a certain way and have a specific function. This is crazy 1 Reply @vishalmishra3046 3 years ago There are 200M proteins and only 137K have known 3D folded structure found using an expensive process of X-ray crystallography. Alpha-Fold2 needs to fill the information gap and publish the results so that the rest of scientific community can confirm the quality of results, by choosing a suitable sub-set and spending the money to validate the results. Reply @elontusk610 3 years ago 7:09 dunno why but this reminded me of the method used to determine the accuracy of the teams who’s job it was to image the black hole a few years back. 2 Reply 8 replies @Jeshua1737 3 years ago It is obvious what environment the amazing Coelacanth, not capable of land locomotion was designed for and its misclassification successfully maintains the consistency of inaccuracy expected. Reply @devonk298 3 years ago well done! 1 Reply @CasperLCat 2 years ago I still haven’t mastered folding my laundry... Reply @leifharmsen 3 years ago I folded plenty of protiens today already. I have done it constantly since my conception. Reply @proteusaugustus 3 years ago I'd like to say your probably the smartest female physicist I've come across. You venture to think outside the box as well. 🇺🇸 Reply @Jim-mn7yq 2 years ago So, how do the amino acids sequenced properly to get the right fold? Reply @DrJBHead 3 years ago So what if it has? More important is information processing to inform protein manufacture and self- organisation Reply @ronimusprime6798 3 years ago I prefer the bottom up approach. Reply @ricardvaletaibanez7193 3 years ago Thank you, beautiful Sabine Reply @Rolandas0R 3 years ago I love this knowledge. Anyway, it is not absolute. Some protein 3D structures cannot be determined because they are never crystalize, like membrane proteins. There were attempts to determine the 3D structure in water, but as far as I know it never succeeded. 20 years ago, when I was a biochemistry student I had an idea how to find this out, but at that time there were no computer power to prove my idea. And no one can do that now as well. The A.I. algorithm is simple: imagine, each atom has some color in light spectrum (quantum color), each atom vibrates and has its own color. Try to combine them all to get the white color. That color is a balance. When molecules become bigger, they need to combine their quantum colors to get the "white" again. It's impossible for the human mind. But A.I. could guess it. Humans only have to give it a formula of thinking. Like mathematics of more than 3 operators: -1+1=0. In three colors we have R+G+B=White. Also, we have 7 colors in rainbow, when you add 7 colors and still get the same white color. A.I. can not learn anything more, or it can do it without knowing us how it did, if we do not provide the A.I. with the correct model of thinking. Reply @loremipsum7ac 3 years ago And how might quantum computing change the protein folding field? Would it be back to top-down approach? Reply @ChargedTTq 3 years ago (edited) The ability to create synthetic bioactive proteins terrifies me... Even compared to the danger crispr poses. From having the ability of a mistake or malicious creation of a contagious and replicating prion. If you've ever looked into prions you know just how scary they are, and how difficult they are to fight. Luckily there are exceptionally few of them that affect humans. They aren't even life forms! They're just molecules that happen to behave like viruses. That lack of complexity makes them very very difficult to fight. At least crispr tools are not contagious. They can feasibly design contagious pathogens but are not in and of themselves dangerous unless directly injected. Reply @ikoukas 3 years ago not all aspects of it have been solved but i am sure the alphafold engineers are working on them too and in five years they will have made great progress Reply @thryce82 3 years ago nice vid. for context (biochem bs learning ml in masters so def not an expert) most proteins cant be solve by crystalogrpahy. The really important parts of the protein are the IDR intrinsically disordered regions. These are the lil bands that that contect the large helices in the structures . They dont take a crystallizable shape. my understanding of crystalography is that it detects repeating structures. indivudual protien collaesce into a crystal those structures that repeat are the ones you can calculate on. IDR are dynamic. they take many no repaeating shapes. thus their structures are hazy to normal techniques. IDR are super megaaaaaa level 9000 imp. they are what allows proteins and their complexes to be dynamic. so without giving intution on those then yeah. 100% not solved. but it seems they have gotten verry verry good at solving the limited structures we have to train on. I hope the model is interpretable and that it can give some intution on IDR cases but tbh the level of compute needed in these sims is ridiccccccuuulous. like a de novo dim on a decent size protein using rosetta (from what ive read) requires a decent amount of time on a supercomputer to even be considered possible. like you need a full NIH grant just to try.......... so doubt it makes much sense other than a black box method :( hope this helps if u took the time to read it ;) Reply @pyr8878 3 years ago Brilliance in presentation, as usual Sabine. So cool, Pluto requests a coat! Love her intellectual passion for understanding the universe as presented to our woefully fleeting human condition! Her unquenchable thirst for knowledge is truly inspirational! d>_0b Reply @Entropy3ko 3 years ago (edited) Don't forget some proteins need "protein chaperones" to properly fold into a functional configuration as well, i.e. not all proteins fold spontaneously into the final functional configuration. - hence it's even more complex. Also people use NMR (usually 2D-NMR) to get protein structure as well. They get triggered if you do not remember them hehe. In any case, AI and automation can only help scientists do their job, rather than replacing them. Reply @1MinuteFlipDoc 3 years ago you are amazing Sabine and I would like to listen to you speak over dinner and a glass of wine. :) Reply @ihbarddx 3 years ago (edited) You don't have to have every protein in your dataset for deep learning to work; however, new predictions for new proteins are, in some sense, extrapolations - less certain. Such methods are the basis of machine translation. The method need not be trained with every phrase in every language, and the result is typically usable, if imperfect. Reply @stefanschnabel2769 3 years ago I think the term "oligomer" refers to several protein molecules interacting. 1 Reply @mattosborne2935 3 years ago Protein, protein, protein, protein We're folding you up, shaping you to plan Protein, protein, protein protein We're folding you up just because we can Reply @Kotsonne 3 years ago proteine folding lyrics: Hey, touch my foldy flaps Grab my terryfolds Grab my foldy holds Grab my terry flaps In my terryfolds Grab my terry flaps You gotta touch 'em... my terryfolds Reply @1337w0n 1 year ago I fold so many proteins every day that I can't even count them all. Reply @jaimecalderonfigueroa1957 3 years ago I understand how predicting the stable state of the protein can be the main goal, but does the algorithm (maybe not the best term in this context) say anything about the mechanism? We already know that the folding cannot go through random search, and that in general there is no unique pathway; so, the folding process itself is quite intriguing. Hopefully the program sheds some light on this as well. 1 Reply 2 replies @FredMontier 2 years ago "... vacum in strings theory" tongue in... Well fone !!! Reply @MolecularArts 3 years ago I solve protein structures by X-ray crystallography. And I’m not worried about AlphaFold. It will mean that we can spend more time asking questions and less time (in the lab) trying to answer them. Reply @maximpapusha5032 3 years ago Loved your string theory joke !!!!! 1 Reply @jonathanmintzer8232 3 years ago Can you develop a Feynman diagram for protein folding? Would this even make sense? Reply @bitflogger 3 years ago Aren't proteins created piece-by-piece? Some folding would be expected, "baked in", before the whole protein is finished, and should be easier to predict. A protein produced in reverse order would be expected to have different folding before its production is finished. Reply @reservoirgeophysicist3013 3 years ago Sabine I am a big fan of yours. Reply @jean-pierrearcoragi6313 3 years ago The problem with the AI pattern matching approach is that it gives us the right answer: 42. But of course we don’t really understand how to solve the problem. It is a bit like the Heisenberg uncertain principle... we cannot have the solution to the problem and how to solve the problem in the same universe :-) Reply 1 reply @leonardobrien 3 years ago Timing is important also. As Kenny Rogers stated. Reply @jpopelish 3 years ago You jumped over the part where biology needs the most stable shape. It seems, to me, that biology may have stumbled on ways a protein, that has many nearly the same stability shapes might use one of the quasi-stable shapes to its advantage, by guiding the folding to that useful result. Reply @lvjkahvlwertfg 3 years ago I wonder where we could be now if the computing power wasted on cryptomining would be harnessed in protein shape calculations.. Reply @mickelodiansurname9578 3 years ago (edited) Sabine, did any of your friends that are physicists publish their findings on the top down approach? Even if they found out there's no visibility on a principle from top down, well that's something right? Reply @ranisharoni75 3 years ago (edited) Protein folding is NP complete in general so it's great insight that nature selects sequences that are fast to fold to a unique stable state. I guess that AlphaFold is not as fast and we are not a simulation ;) Reply @MrFishluver 3 years ago I helped with protein folding with my PS3 back in the day. You look MARVELOUS in that black dress from your other video, more please... Reply @lightningcoreeygja9678 3 years ago Excellent. Reply @garymcleod9170 6 months ago Glad you are making biological concepts clear. I watched quite a few physics videos you produced and always wished you would take your presentation skills to biology. Reply @georgesos 3 years ago I loved the joke about string theory ..:) Classic Sabine Reply @digitalconsciousness 3 years ago If medications were developed from this tech / knowledge, what form would it be in? Pills? Injections? Reply @jamielondon6436 3 years ago Sick String Theory burn. ;-) Reply @EstamosDe 3 years ago Thanks you! Reply @jeffreykalb9752 2 years ago The question doesn't even make sense. Protein-folding is not a single problem, but thousands of different problems. Reply @thomasb6573 3 years ago Ulf Grenander's book, General Pattern Theory, should have insight into predicting protein folding patterns. Reply 1 reply @peterwhyte317 3 years ago If you have ever carelessly cut the bindings of a big coil of spring wire you will get a feel for the problem. Reply @Amipotsophspond 3 years ago Not Grumpy because I am not in that field. HOWEVER, I think the data set might be a little small on this contest and we might be seeing some sneaky over-fitting. Reply @rodnorris9532 3 years ago (edited) Very good explanation. But I did not hear you explain why they fold. They assume a shape that allows them to 'mate' with another specific protein. Much like a key in a lock. And that assumes a lot of Natural Intelligence. Reply 1 reply @kaystephan2610 1 year ago The criticism is understandable and I guess also reasonable because as something new arises that appears incredibly awesome, it is best to at first remain sceptical instead of instantly hoping for a miracle and willingly accepting flawed data. HOWEVER: if we look at the jump from 2018 to 2020 we can get a grasp at how quickly this stuff is improving. I don't think AI will replace researchers completely. But AI can be used to reduce the amount of time and effort that's needed to extract information from Data. That saved time and effort can be put into new findings by researchers. Reply @TheMrCougarful 3 years ago LOL throwing shade at String Theory. Reply @TotalRookie_LV 3 years ago Now I wonder how much I've done for the cause by running BOINC on my PCs. Reply @cesarjom 3 years ago ML may provide the correct folding structure for a given protein sequence (amino acid), but unfortunately due the the "back box" nature of ML algorithms and training, biochemistry will still not understand why/how the sequences produce the specific folding -- that is additional needed research to gain that knowledge. Reply @alanguile8945 3 years ago I've folded millions of proteins every. day, life hay isn't it marvellous. Reply @rajeevgangal542 3 years ago AI ML can only learn patterns...hmm. that's exactly what structural biologists have been doing all this while. Folding templates and recurring motifs were always used alongside an understanding of forcefields, molecule binding interaction and cellular environment. The patterns learnt by ML are generally generic enough to be applied to unseen data, provided enough of feature and response space has been sampled. Exactly why they work. Another thing you could've mentioned is the intermediate secondary structure prediction that has been the hallmark of previous methods. Reply @Joe-iv5ks 3 years ago Are the protein folds instructions mapped in the DNA? and then built by Genes? Reply @sjzara 3 years ago I may be one of the grumpy ones, but as I no longer work as a biochemist, I don’t fear AI taking over! I don’t consider what these AIs have achieved as “solving” protein folding. If protein folding had truly been solved we would be able to go from desired shape directly to sequence, but we can’t. We can only go from sequence to approximate shape. This is extremely useful, but it’s not giving us insight into why proteins fold into the shapes that they do. Truly having solved folding would result in us being able to look at a reaction, work out the amino acid positions that would catalyse the reaction, and then go straight to the stable sequence that results in those amino acid positions. We are a very long way from that. Reply @sparkyy0007 3 years ago (edited) The search space for a valid fold sequence according to the 20 year research project cited below (2004), quantified the problem at a single valid fold in 10E77 sequences on a 150 amino polypeptide chain. If this is true, and considering there have only been 10E40 organisms that have ever existed, random mutations and selection cannot explain life, not even close. If first life had 3000 protein varieties, and we have 120,000, most much larger than 150 amino units, 117,000 valid proteins would need to have evolved by chance. 117,000 X 10E77 = 10E(9,090,000) And that's a wildly conservative estimate as most proteins are far larger than 150 units. Not a hope in hell chance mutations and selection could have accomplished this. The prevalence of low-level function in four such experiments indicates that roughly one in 10(64) signature-consistent sequences forms a working domain. Combined with the estimated prevalence of plausible hydropathic patterns (for any fold) and of relevant folds for particular functions, this implies the overall prevalence of sequences performing a specific function by any domain-sized fold may be as low as 1 in 10(77), adding to the body of evidence that functional folds require highly extraordinary sequences. https://pubmed.ncbi.nlm.nih.gov/15321723/ Reply @nicholasmarks4288 3 years ago This is a new discipline for me but it suggests to me that the greatest secret here is learning how the living-cell replicates. My unification theory says that the living cell, in its component parts, is two part atoms that have a place on the periodic table. It is that place immediately before the hydrogen atom. In fact it is an incomplete hydrogen atom that was spewed out of the sun before its higgs-field was able to snatch a proton or electron and millions of them came to planet Earth piggy-backed chemically by their nearest relative...the H atom as they were all evicted during an early solar max. It is the energy generated by the coming together of these massless part-atoms that grabs hold of atoms and swirls them around in their dna structures and it is this, I suggest, that determines the folding of proteins according to the electric strength permitted for the purpose...which can vary according to the respect given to our nervous health. Reply @adamburling9551 3 years ago Interesting. Although they'll never even begin to understand a cell or get anywhere close to creating one. Reply @BjornMoren 3 years ago Excuse my biology ignorance, but I thought all proteins were assembled amino acid by amino acid, much like an assembly line. So doesn't the protein immediately start folding as it comes out of the ribosome? So is it possible to break down the problem into small steps, like "how does the next attached amino acid make the protein fold, etc?" Reply @robicjedi 3 years ago It is not true that you need actual measurements to train AI. As rightly pointed out, one can calculate if one configuration is stable, there are just way too many to check. The trick here is that the algorithm will explore some random possibilities and then learn which possibilities to check next with a high success rate based on prior experience. This way not all possibilities need to be explored. Deep neural networks (DNN) just store the experience of the algorithm, but they themselves do not carry out the exploration. The actual technique used is reinforcement learning, which has many algorithms, one of which is deep reinforcement learning. Reply @danwat1234 1 year ago Protein Folding! I am aprox. #10 top contributor on the Rosetta at home Distributed Computing project Reply @i486DX66 3 years ago The goal is to be rendered obsolete by software. Frees people up to work on other problems. Reply @falkoskidaniel 3 years ago Fucking hell...How could you explain a complex topic so well??? It is such a amazing video. Thanks a lot for sharing it . Reply @user255 3 years ago Misfolding seems to be one of the pathological mechanisms of type 2 diabetes, but its cause is in lifestyle. Not in protein folding. Reply @mojeimja 3 years ago If we only recently reached 60% success, what was Folding@HOME really doing ? 1 Reply @WilliamGacquer 3 years ago Did not know that NordVPN was involved in Protein Folding. Reply @georgeorwell8138 3 years ago Love Sabine Reply @Levon9404 3 years ago To me you are the smartest woman in the world Sabine! Reply @NeoKailthas 3 years ago These people need to get 90% then I will listen to their complaints Reply @Petrov3434 3 years ago Dear Dr. Hossenfelder (I am uncomfortable calling you more intimately Sabine - we are both European), Have a question on Diabetes-2 caused -- by improper folding of protein... Obese people when have "stomach stapling procedure" lose weight and -- no longer diabetes-2 is present (no need for insulin at all) My dentist is one, among many, who is a living proof -- he was morbidly obese and now has no sign of diabetes... Do you have hypothesis what happened? Many thanks in advance 1 Reply @Hei1Bao4 3 years ago I've been testing the waters of proteomics and quantum biology lately, and keep finding myself confused by certain definitions, such that of "an aperiodic crystal" in reference to DNA. "An aperiodic crystal is a structure with sharp diffraction peaks, but without lattice periodicity." Parsing that definition sent me down the rabbit hole. Think you could tackle a video on that? Thanks so much for your clear and concise explanations! Reply 4 replies @sapelesteve 3 years ago Very interesting topic Sabine. Now I am wondering how a molecule like DNA, which is composed of large numbers of amino acids, always has the double helix shape. Anyway, great video from you as always! Reply 4 replies @TheTractorjaws 3 years ago I could listen to you all night long! Reply @Handelsbilanzdefizit 3 years ago (edited) Maybe, Proteins that "replace" plastic would be better. AI is very good in such things, like finding patterns, structures, relations in a lot of data. But there's also a domain where AI is very bad. Training these models do not only require a lot of data, often it requires a lot of energy. When the model is trained, it's a static function (input --> output), and can not adapt any more. For an update, you often have to train it from the beginning to make it fully reliable. In some tasks, AI has very good results, but is very inefficient too. So, it's like a space rocket. You have to accept 95% fuelmass an 2% cargo, because it's the only practical technology to reach space. There's no other way. AI can not think and reason, so it won't find complex mathematical proofs. And similar to quantum-computers the output has some little error (Histogram distribution). So, overall, I would compare A.I. with the euphoric promises of nuclear-technology at the 60s. AI has it's domains where it performs well, but we will see downsides in future. Reply @johnnysparkleface3096 3 years ago (edited) Can one be made that can unfold other proteins? Reply @nicolasduguay4 3 years ago My dad is dying of ALS, My grand-father is dying of alzheimer, I never though protein folding could be that important to me! 1 Reply 2 replies @Furiends 3 years ago It's hard to say when AI will go but it really can't be overstated how meaningless 90% is to "solving" protein folding. It just shows us progress has been made. Being wrong 10% of the time is okay when the alternative gets it wrong more often. This is the benefit with certain applications of AI like autonomous driving. However simply having a road system without traffic conflicts would avoid the need for AI in the first place. How hard is that to build anyway? Reply @stevedyches4635 3 years ago So alluring this woman is. Reply @saadmahboob3084 3 years ago Great content. Reply @oznerriznick2474 3 years ago Great video. 10 to the 50th power possibilities? That's not very much...hilarious.. Reply @Lemonz1989 3 years ago It would be pretty cool if robots could "invent" or find out new information themselves, without needing to 100% rely on already known information. Would be amazing if robots and scientists would invent/research things together, like how (human) scientific collaborations work now. :) Reply @hillaryclinton2415 3 years ago So, throw a fourier transform at it. Or is it the other way around? Reply @alexjaybrady 3 years ago An organism that can eat plastics sounds like an apocalyptic movie plot where its called the Plastic Plague or something! Reply @unnamedchannel1237 3 years ago No idea what this is all about but it was funny as hell when Andre the giant Took a poop on bad news brown during a ring match Reply @i.m.i.7310 3 years ago Comic relief to the rescue penumbra and jelly sammich. Lol. Day off my pleasure as always your genius is obvious to me.. danke fraunhofer research.... Sabine is a luminous star. ☆ Reply @alexojideagu 3 years ago I love the bio mechanical " Machines" inside life. 1 Reply @rikimitchell916 3 years ago Sabine though the possible number of folding algorithms maybe astronomical the probable number is much less due to chirality and dipolar electrostatic repulsion Reply 1 reply @gianfrancofronzi8368 3 years ago Good thing to find out what is up with the protein folding. I didn't know about it until I saw this video. But it has to do with all biological life in a way because protein is a building block of, I believe all life, and it probably takes in a big part of a problem that we would like to solve. Considering that, it is the basis of diseases that, I assume could lead to a bigger correction that could continue onwards by correction at the start ? Reply @bradhilton2283 2 years ago Alpha fold is really just in its infancy, and it’s all ready solving the “‘Casp “ companion at 90% imagine Alpha fold 7 ? The applications would mostly likely be immeasurable . Reply @jan-martinulvag1953 3 years ago I am ( my thoughts) the environment of my proteins. 2 Reply @julienyt1600 2 years ago Also AI is it's own field of research, and we are looking for better ways to protect against out of sample data using generative networks and other exploratory algorithms. Reply @johndavis6119 3 years ago Still folding proteins on my computer when I’m not using it. Reply @jaxxonbalboa3243 3 years ago I folded time & space to get here to watch folding protein. 1 Reply @imjustpassinthru7779 3 years ago When the proteins that living organisms are made of mutate, what percentage of mutations are in any way useful to an organism? I have read that the ratio of useful mutations to non-functional or harmful mutations is only one in ten to the 77th power Reply 1 reply @MBS_Drew 3 years ago This still doesn't explain how proteins begin folding AS they are being constructed. Meaning, they somehow know the configuration they will become before they have all the amino acids that would determine the structure. Reply @lukaspfitscher8737 3 years ago (edited) Why not using Gradient descent, the form with the lowest energy Reply @pppdddqqqbbb 2 years ago It went quite all ok till you've reached VPN adv., then there the b/s began. VPN has nothing to do with safety *at all*, it's a matter of ISP changing — instead of physical connection to his Internet Service Provider one gets connected to a VPN provider in the first place. Reply @michaelstora70 3 years ago This is the true origin of Slagathor in 1988. Not Scrubs, not Venture Brothers. Reply @hanscyrus 2 years ago All proteins begin to "grow" towards functional maturity AND then begin to fold, all in some type of specific environment unique to their own. Perhaps begin to think to model and simulate these protein environments and towards why each protein is folded as such? Reply @mattkerle81 3 years ago How is type two diabetes related to protein folding? I thought that was due to metabolic syndrome, did you mean type one? 1 Reply @Septumsempra8818 3 years ago (edited) Top down simulated in the sequence they leave ribosome will beat AI Reply @ThorkilKowalski 3 years ago You are amazing! Reply @mvmv-pn8zt 3 years ago If the number of protein folds is so vast how do you discern that proteins are the wrong shape and causing harm? Reply @MichaelHarrisIreland 3 years ago This gives me a break from the crazy world. ....from Ireland. Reply @siamakafrasiabi7198 3 years ago You are super :) Thank you so much. Reply @JohnBoen 3 years ago I'd like to see a video on why transformers are so awesome. ...and why we should all expect to be stunned multiple times a year from now on. For example, last year we also saw GPT-3 results. From this, I think we can see the end of mainstream software developers, short topic authors, and many other creative type jobs in about a decade. Can you put a positive spin on 50 million US jobs going away in 25 years? Reply @virtualrealitychannel2276 3 years ago Is this where I can get the latest protein folding news? Reply @eljcd 3 years ago There is an interesting article at quantamagazine: Some Proteins Change Their Folds to Perform Different Jobs Looks like the problem gets more interesting... Reply @vgernyc 3 years ago Skynet will design Terminator Prions Reply @PythonPlusPlus 1 year ago The idea that AI can’t come up with solutions for new problems is silly. The reason AI is so powerful is that it’s capable of learning the abstract concepts that allow you to come to a solution. What it learns from folding current known proteins can be applied to newer proteins that we haven’t discovered the structure of yet. To top this off, 90% is only the current accuracy. The AI research hasn’t stopped, Deep Mind will continue trying to improve their AI until it is 100% accurate. Reply @rseyedoc 3 years ago Do proteins fold in states of superposition? Reply @einsteindrieu 3 years ago Need to know more on the electron and the atom how many times the electron goes around the nucleus per second. Reply 3 replies @ericvelasquez1282 2 years ago I'm surprise that the origami masters has not solve the protein folding problem Reply @PetrGladkikh 3 years ago 4:00 In hindsight everyone is smart, but still: to me it seems natural that number of possible folded configurations is limited or even 1, because the aminoacid sequence is synthesised _sequentially_, from start to end. And even if it is only several bases long yet - it starts fiolding. It is not like the video that is included at 1:35, no one holds the sequence stretched until it is completely done. I also wonder why when they describe folding modelling this never gets attention. Or is it actually known to have no effect? Reply 1 reply @mikepict9011 3 years ago Probably pretty important for living in low gravity too . Reply @cyclingadventure1804 3 years ago You are fabulous! Reply @BillyMcBride 3 years ago (edited) Can one improperly folded protein molecule cause illness? Reply @lodrezzon 2 years ago AI will continue to improve, as will robotics, and robotic aesthetics. If we don't kill ourselves over land grabs and wealth hoarding, the future will be an amazing place to live, with humans and artificially intelligent machines working hand in hand to make the world a better place. I don't share the doom and gloom ideology that AI can only lead to humanity's destruction or widescale poverty. Political initiatives must be put in place to distribute technological innovations to all, so all can benefit from these inventions. I think these initiatives will be more challenging than actually creating the machines, given mankind's penchant for personal gain at the expense of social development and prosperity . But, developed correctly, it will be the greatest invention mankind has ever created in helping man to solve unbelievably difficult problems in short periods of time, leading to a truly utopian future. If we can just learn to overcome the single greatest hurdle to humanity, our own shortcomings. 2 Reply @flugschulerfluglehrer7139 2 years ago Bottom up or down? Deduction, Reduction, Induction. Reply @jaidevsharma570 2 years ago i just have one question. why we don't discuss about ramachandran plot. when it comes to protein folding Reply @BartJBols 3 years ago Ive actually folded proteins in real life, in fact, im doing it right now, while watching this video. Reply @peterclark6290 3 years ago Some real Science issues at last, using actual data, peer-review-like processes. Glad to see the Mathematics moved back to level two, where it permanently belongs. Reply @johnmcdevitt3805 3 years ago I’m going to have a medium rare T-bone steak . Wonder if it matters how it’s cooked to get the maximum protein benefit. John McDevitt - The Retired guy Reply @ThomasJr 2 years ago The audio of this video is off. Someone messed up. Reply @Wallach_a 3 years ago Still waiting for Sabine to say “dooblydoo” Reply @anneallison6402 3 years ago Are you available for interview? I work for a radio program Reply @jeandauvergne6957 3 years ago (edited) Protein folding and above all 3D structures predictions will certainly be achieved by the AI. As you´re saying, training and accumulating data are necessary, but the huge increasing possibilities of quantum computers will completely change this area of research. Reply @mattiefee 3 years ago I knew my doctorate in origami would come in handy! Reply @blasater 3 years ago (edited) Interesting but it still doesnt explain the other proteins that are required to fold proteins into their native states called chaperonin. How does a chaperone appear on the scene and how does it know which protein it needs to fold? It is a mind-blowing problem. Reply @markspqr 3 years ago She is the absolute best ... Reply @piggypiggypig1746 3 years ago encouraging Reply @michaeldebellis4202 1 year ago Isn’t there another issue though? Just predicting is not a theory. The Deep Mind program may be great at predicting but it still doesn’t tell us WHY those forms and not others. To me it seems analogous to feeding a bunch of data into a program about moving objects so for any new object if you know the force and the mass you can reliably predict the acceleration. That’s useful but it isn’t the same as knowing that F = ma. Or perhaps a better example would be predicting orbits and trajectories of objects in space. Being able to predict via analyzing data isn’t the same as having a theory that gravity is bending of space-time. Reply @scottfranco1962 3 years ago So I always wondered how DNA can be so long and be in a compact space without getting tangled. If I put my earphones in my pocket, they come out so tangled it takes a half hour to untangle. Turns out the answer is: nature has that problem as well. DNA is kept in a special structure that keeps that from happening. 1 Reply @szolanek 3 years ago (edited) I like Miss Wikipedia! Really. Some commenters said, she started with music and followed with science. .... Hmm, the two things are the same. Only approaches from a different angle. * Also, Brits ran AlphaPol for predicting the next move of Boris Johnson. It burned down. Reply @janwaska4081 3 years ago @ around 3:30 Sabine said something really funny: That’s almost as many [folding] variants as there are string theories! That was really funny, though perhaps some of her fellow physicists were not too amused by her excellent joke. Oh, well. Too bad. ;) Reply @robertweekes5783 1 year ago They’ll probably get the accuracy to 97% and then 99% the year after Reply @giveaway4002 3 years ago interesting video 1 Reply @bhangrafan4480 3 years ago The sad thing about the AI approach, (apart from the fact that it is limited by the contents of the training set), is that it gives no insights into the behaviour of protein molecules which might give a greater understanding of how proteins actually function. Although applications such as drug discovery can get a big boost from this, (if it is sufficiently reliable), the academic is left feeling cheated. We gain no knowledge about the character and behaviour of the molecule which is what the academic really wants. This is not a purely academic issue, such understanding could enable us to design artificial proteins with useful properties. 1 Reply 2 replies @Misiulo 3 years ago We need a programmable nanite, capable of folding protein physically. Reply @christopherrubicam4474 3 years ago Ha! Biochemistry explained by a physicist. Thanks, Sabine. Reply @el_engineer0p106 3 years ago i love her voice. Reply @juliane__ 2 years ago Very glad, i could contribut weeks of IT Power to Fold IT. Reply @zeroqp 2 years ago Abbreviations use a punctuation to indicate a shortened letter. Hence, "artificial intelligence" is "A.I.", not "A.I". Reply @millicentduke6652 3 years ago There’s a tiny voice in the back of my head telling me to remember that Sabine might be an artificial intelligence making content to reassure humans of AI’s benign or even symbiotic relationship to us. 1 Reply @frankdavidson644 3 years ago Amazing Reply @blackpoolbootz2790 3 years ago Interesting. Remember was a protein folding program could run on a pc in background to use computer resources along with hundreds of other users. Reply @nigeldepledge3790 3 years ago This sounds like a very impressive achievement. Protein folding is an immense and complex subject, and you've distilled out the key points very well. One application for which a fundamental understanding of protein folding would be extremely valuable is in the realm of biopharmaceuticals. If your target protein is produced in a culture of mammalian cells, then it is likely to be correctly folded but purifying it from all the thousands of proteins that mammalian cells also produce for their maintenance and metabolism can be a big challenge. If your protein is produced in a microbial expression system, then it is far less likely to be correctly folded. Indeed, mammalian proteins expressed in microbes often aggregate into amorphous blobs known as inclusion bodies. It's easier to purify the protein in this form, but you must then refold it before it can be useful. Reply 7 replies @03chrisv 3 years ago (edited) I remember when people were running protein folding software on their PS3 consoles. Reply @kateekat 3 years ago Informative! Algorithm boost 🇺🇸 Reply @LiborTinka 3 years ago Quantum simulator, i.e. quantum computer would solve it. This is exactly the type of problem outside classical computing. Reply @danniles5256 2 years ago (edited) A lot of PHDs are very excited about this....also i love your channel...you are so cute. Reply @geralldus 1 year ago But… how did such stable and complex folding originate in the first place. It could hardly be random, with probabilities of the power of 300 the universe hardly seems old enough or am I failing to understand the whole concept? Reply @0x0404 3 years ago Alpha Go was 2015? It feels so recent. Reply @chookvalve 3 years ago Is there an English version of this? Reply @maxsharifi3803 3 years ago Nice! Reply @swapanghosh4880 3 years ago Could have promising ablity to edit genetic defects to eliminate genetic disease and infirmities due to things like arsenic in water and toxins in mustard oil specifically one locally called dhatura. Reply @holden.a 3 years ago I think it is not prediction! The software learns what shape belongs to a sequence, because it gets the data as input. Then it does not predict, only tells what it has got from the input data. I'd like to know what is the input data like, which is put to the neural network. If it contains the sequences and the shapes then it is only a fraud, the software gives it back what it has got, it does not predict anything. Reply @NephilimFree 6 months ago The smallest protein known is a 20 amino acid protein called Trp-cage in the saliva of Gila Monsters. The smallest human protein is 44 amino acids but it could be an abortive translation from the 5' UTR of another mRNA. The largest known protein is the human muscle binding protein Titan, which has btween 25,000 and 35,000 amino acids depending u8pon the reading frame for the sequence. There are no proteins known which as so small that they are comprised of only 10 amino acids. A typical protein is 300 - 3,000 amino acids long, while thousands of proteins are many hundreds to thousands of amino acids long, and some are tens of thousands of amino acids long. For such a protein to form naturally into a functional form can be calculated. There are 20 amino acids that make up all protiens. Therefore the odds to form a protein comprised of only 10 amino acids (no protein this small is known) by undirected molecular interactions that is functional is 20x 20 x 20 x 20 x 20 x 20 x 20 x 20 x 20 x 20 = 1:10,240,800,000,000. (240 trillion 800 billion). Reply @aniksamiurrahman6365 3 years ago My goodness, where did you learn that much Biochemistry? Reply @tabularasa0606 3 years ago It's not entirely true that deep neural nets can only solve problems they have been trained with. AlphaGo made a few surprising new moves good that no professional player had made before. It created a new pattern. Reply @RandyJames22 3 years ago When it matures, how quickly can quantum computing push the envelope on our knowledge of protein folding? Reply @johnmcdevitt3805 2 years ago Very sharp mind Reply @jambec144 3 years ago (edited) @ 7:52 Is this "A.I" the same thing as "A.I."? Or is it analogous to "A.you"? Reply @The0ldg0at 3 years ago Has any scientist checked the effect of phonons and resonant frequencies on protein folding? Reply @jamesr2408 1 year ago This comes down to proving that the basis of life is not not the silly Darwinian chance but super and supreme intelligence driven design. Reply @OnTheRiver66 3 years ago This is encouraging. Some day we might have a cure for diseases caused by prions. Reply @mikejones-vd3fg 3 years ago what a time to be a alive! Reply @musicalBurr 3 years ago (edited) The snap zooms on the cuts is mostly more distracting than simple jump cuts. If I were you I’d lose that effect, or at least use it more judiciously, like when an important new point is being made, then use a snap zoom. Keep many jump cuts - it’s part of the new vblog cinematic language, people accept it.. Just weighing in for what it’s worth (I know you have a new editor etc.) Reply @charlesbrightman4237 3 years ago 3 Basic Energy Forces In This Universe: Gravity, Electrical, Magnetic. Energy interacting with itself, causes things to occur. Reply @aljodomo 3 years ago Laptop or phone? I use a tower pc. Am I not allowed to use NordVPN? :( Reply @joaquinel 2 years ago Alexander solved a complex one centuries ago... Reply @LaurentLaborde 3 years ago 10mn to go to the part that interested me then, message from the sponsors and end of video. dang... :( Reply @mmicoski 3 years ago Sabine: Cyrus Levinthal estimated a typical protein could fold in 10^140 ways Me: what??!! :o Sabine: don't take this number too seriously... small proteins fold in "only" 10^50 ways. Me: phew, now we are talking. Let me take my notebook ;) Reply @kricsek 3 years ago So is Folding@Home useful? Reply @cillian_scott 3 years ago Type 2 diabetes caused by protein misfolding? Which protein(s)? Reply @homejonny9326 3 years ago I think the most viable solution to this problem seems to do a lot of trial and errors tests. In the future, the amount of positive cases database will be enough for most cases, and keeping always increasing, so... Reply @youdodat2 3 years ago Not GOOGLE! DAMN IT! Reply @markphc99 3 years ago i thought we were waiting for super duper quantum computers to sort out foldings Reply @SquizzMe 3 years ago (edited) That String Theory shade though.... Drag her sis. Reply @baljeetbhachu4273 3 years ago could tesla's dojo be used for folding Reply @sandrasandra7593 3 years ago Thank you so much for sharing science and so making a better world, really enhancing democracy Reply @Boyer427 3 years ago How did Natural Selection figure out how the linear DNA mange to know how the 3D protein would fold into correct form, with all the catalytic surfaces of the protein in their respectively correct positions, unless time did not exist or the folded protein was a phase of time Reply @happyfaceonU9 3 years ago Can you do a video on dna mutations? Reply @mydogbrian4814 3 years ago - With so many possibilities (1×10¹³⁰) in just protien foldings alone & with nearly all of them bogus to the biological mechanism. Does our concept of spontaneous generation of life from inert matter even make any rational sence? (e.g. Life on Mars). Reply @kthwkr 3 years ago vacua = ? A new word to me and after looking it up I still don't understand the use of that word. Reply @nicka.papanikolaou9475 1 year ago Sabine, the answer is no it has not been solved, not by any stretch. AlphaFold's as exciting as it is, it is based on known examples, and we have thousands.So, there is no de novo prediction yet. Reply @technofeeling2462 3 years ago I do it from time to time when I put wurst on my bread Reply @BillyMcBride 3 years ago (edited) If you think of spacetime as an algorithm which we have in our minds a priori, why can't we find that groove in us where we are just as good if not better than computers at learning and solving with our intuition like problems? Reply 2 replies @malekmannai9445 3 years ago That's almost as many vacua as there are in string theory 2 Reply @666Metalbassist 3 years ago Oooooohhhh as a biochemist by trade this intrigues me Reply @workin4alivin585 3 years ago (edited) Evil is the lack of a good that should be there. Disease. Misfolded protiens. Reply @lorenh763 3 years ago They've had protein wraps for years.... how is that not protein folding? Reply @janerussell3472 3 years ago (edited) Everything decays. Even as we develop from the seed, we decay. Just as physicists talk in terms of symmetry-breaking, life can be thought of in terms of cell division. The remarkable thing is that every cell contains the complete nuclear DNA code, ( aside from the red blood cells.) If everything is running down, logic dictates it was wound up. We probably started with one chromosome, like the transcendental jumper ant [ they all strobe together ] and no "junk" genes [ we now have some 98 or 99% non-coding. ] Just like asking where a Big Bang came from, how was life's clock wound? Darwin was looking through the wrong end of the telescope, in my opinion. Is mutation going to be beneficial, as a rule? Is a serial process on a random walk going to produce functional complexity? In the real world, away from fantasy physics and evotards, that requires an intelligent designer. Reply @dogzer 3 years ago Never heard of protein folding before lol Reply @vishalpachpande5921 2 years ago The problem with ai you mentioned is not actually AI problem ....it's deep learning problem....AI is vast field and there are other methods that does not rely on external data. Reply @lenny108 3 years ago Protein folding is so complex that it is difficult to assume that this can happen by random mutation. Reply 1 reply @mysteryguest9555 3 years ago 10 to the 300 combinations... jesh Reply @dr4t 3 years ago Back in my days, proteins were being folded by hand Reply @srobertweiser 3 years ago My grandma used to be great at origami, I’ll bet she could’ve folded them damn protein. Now, she couldn’t fold a beach towel, she’s crazier than a shit house rat. Reply @nealinnc 3 years ago God figured it out... Reply @lisaschuster686 2 years ago Crowd sourcing! Reply @mizalamizala2317 3 years ago Waaay over my head.. Reply @laurum1318 3 years ago Ok Sabine, you are awesome because you are a physicist woman and everything you say is very interesting. I also love your outfits!!! Consequently, you are my role model ️ Reply @danrowley6934 3 years ago Sabine (You rock) Reply @MCsCreations 3 years ago Really, really interesting!!! Well, they already got it about hormones, years ago. But it's made by engineered bacteria... So I don't know how it would apply. 1 Reply 4 replies @nomorestealing7152 3 years ago Let's keep it really simple it is minerals that determine protien foldind you know shuch as sufur , 3disufide bonds in the protien insulin. Its called metaoprotiomics Reply @jonathanmoore4837 3 years ago Wow, NordVPN looks very useful Reply @christopherconroy2805 3 years ago Good Communicator! Reply @Knight766 3 years ago You're the best Reply @faustin289 3 years ago Woow, I remember how DeepMind's AlphaGo and Zero were hyped back then. How come then the media kept completely silent about AlphaFold? At least I would have had something to cheer me up about the damned year of 2020. Reply 1 reply @oker59 3 years ago The protein folding problem has not been solved mathematically. There's no deductive proof for a closed form solution to the protein folding problem. Alphafold has shown they can make good approximations to be able to do something technologically. There will be improvements. The computer they used was not the fastest computer one could get their hands on. I've emailed them about a Cerebras wafer scale chip that has one trillion transistors. I've also emailed them about using protein folding to make nanotechnology happen. Eric Drexler tried to suggest proteins as a way to making nano-manufacturing happen. The ability to make things the way life does; grow things instead of chip away. Well, he knew of the protein folding problem and tried to suggest designing proteins to be more predictable. I guess that didn't work. But, now we've been able to predict, with alphafold, pretty complex proteins. And, improvements will be made to be able to predict ever more complex proteins. But, nanotechnology probably doesn't need these complex proteins. They need simpler more uniform proteins I would think. I'm thinking alphafold knows this. I've seen some evidence of others trying to get alphafold to try to make nanotechnology happen. And, foresight institute has been dead silent for over a month now. Almost as soon as the alphafold annoucement was made, the foresight institute has gone very quiet! Reply @Marshaluranus 3 years ago roflamo that stein guy looks like dr. strangelove Reply @Chris.Davies 3 years ago (edited) Betteridge's Law of Headlines says NO. But, let's watch to see if the law is correct... Yep - Betteridge is NEVER wrong, and using closed questions in your titles seems like lazy clickbait. I think you are above such things, and voted this video up, despite violating Betteridge's Law. A much better title would be "Where are we at with the difficult Protein Folding problem?" That's an open-ended question, and it means you're going to tell us all about it! One final point I'd like to make. Human languages all feature the same thing: a period, or fullstop at the end of each sentence. The period is important, because it ends a single oral presentation, and allows the speaker to draw a breath, and then continue. The time it takes to draw that breath and begin the next sentence is also the time it takes for a listener to process the sentence, and prepare to listen to the next sentence. If you make a cut at the end of every sentence and do not leave in the time taken to draw a new breath (around 500ms to 1 second) before you cut in the next sentence, then you have removed our ability to properly listen to you, and to properly comprehend what you are saying. The efficiency of communication is actually INCREASED by increasing the gap between sentences, and that is why the best speakers in the world leave a conscious and purposeful gap between sentences, as they know it adds impact to every sentence they speak. A good speaker is NOT afraid of the silent pause between sentences, and uses it profitably, instead of editing it out in post-production! Take a look at the despicable PBS Eons and SciShow YouTube channels where they insult us with a constant stream of consciousness without any break at all, which leaves us feeling abused, and stupid, and not properly educated by whatever info they are trying (and failing) to present. Thank you, Sabine. I enjoy your videos greatly and would enjoy them even more if you left the natural pauses in your sentences. 1 Reply @LeeRaldar 3 years ago As an ex-chef my main experience with proteins was in unfolding them. Reply @klausgartenstiel4586 3 years ago a.i. increases the life expectancy of old people. but it has useful applications as well... so i'm on the fence. Reply @martinfederico7269 3 years ago You did not mention the essential role that chaperones have in protein folding, which is not included in these models Reply @sodalitia 3 years ago Take a protein shake shot, every time she says protein. 1 Reply 1 reply @shqueeblesmckloofin7351 3 years ago Check out Alyssa Hill from OU, Norman. Reply @Raydensheraj 3 years ago I think you and Sean Carroll need to have a talk. 1 Reply 1 reply @AkarshanSarkar 3 years ago Sabine is so persuasive , I got the NordVPN right after this. (PS I was looking for a VPN for a long time anyway :-P) 1 Reply 4 replies @BlackHermit 3 years ago It is not the End, of course, but scientific progress is important nonetheless. Reply @joaodecarvalho7012 3 years ago "Go will never be mastered by computers." 2 Reply 8 replies @JanneWolterbeek 3 years ago Oops, I already have NordVPN! Thanks for the insightful video though! Reply @z0000g7 3 years ago What about chaperone proteins that shape other proteins by inducing them to fold a certain way as they're being made, in ways that would be unlikely otherwise strictly based on their amino acid sequence? If many biologically useful protein structures are made this way, then surely protein folding CAN'T be down to amino acid sequence alone? Reply 3 replies @Yellowblam 3 years ago Great news. Reply @eljcd 3 years ago Well, first I'll get this out: terms like "deep learning" and deep mind" creep me out. Sounds as if the machines will think for us now. About he video, my years of Biology's student are well beyond me, but I think I remember that there exist "stevedore" enzimes that guide the correct folding of the proteins, and the celular environment intervenes too, for example by the existent pH. One wonders how much of this is taken in account in these simulations. 2 Reply 2 replies @MrHenkfromHolland 1 year ago I’m a member of folding@home Reply @TheDoomWizard 3 years ago Fingers crossed Reply @Qoow8e1deDgikQ9m3ZG 3 years ago don't blame too much.... F=ma is very good indeed.... 90% is not a bad score. Reply @colfaxschuyler3675 3 years ago I am still left with the feeling "so what?". What does the knowledge lead to? Is it used to make folded proteins? Given the size, how many would be needed? How are they made? What is done with them? If we sent code or to everybody's cell phones, and spent a year crowd-sourcing folding, what works the information be used for? Reply 2 replies @dsc4178 3 years ago Shape determines function. Proteins know this. And endoplasmic reticulums, Golgi's.... Reply @Neuralatrophy 3 years ago The problem with using AI to do it is it's not REALLY solved. The task may be reliable but WE don't know exactly what the AI is using to predict the folds. Reply @miklov 3 years ago I find it weird that deep neural networks are called artificial intelligence, do they show signs of intelligence? To me they seem to just be big stacks of layered non linear functions that are kinda tricky to have good control over in terms of measuring the quality of the predictions. Maybe I just have a weird idea what artificial intelligence is ^^ 1 Reply @peterolbrisch1653 3 years ago I can't even manage to fold my clothes. Reply @wiadroman 3 years ago Problem of protein folding is nothing in comparison to the problem of my t-shirts folding. Reply @Eazpezey 3 years ago The last 10 percent can take 1000 years, 1 percent per decade...omg Reply @ianedmonds9191 1 year ago Sabine always comes across a bit cynical but I've come to realise this is a germanic trait and not to be misunderstood as being discouraging of the innovation being talked about. Source. I have a boss who grew up in Leipzig. Reply 1 reply @mebossyounothing 3 years ago Proteins are C++ Molecules are Assembler Atoms are Binary machine language Reply @ashirahelat4749 2 years ago You have pretty eyes Love this topic Reply @etbadaboum 3 years ago ...Vacua? Reply @jhoughjr1 3 years ago im going to guess no Reply @onlymeok 3 years ago 10^140 foldings. The Thing confirmed. Reply @dogwithamug 3 years ago That string theory own lol Reply @Franciscasieri 3 years ago In 200 years, we will go to the doctor to get medicine to prevent inherited folding mutations. Reply @SimonBuchanNz 3 years ago AI only being able to predict based on what it's been trained on is a bit of a weak argument: that describes every approach after all! Probably the intent is that neural networks don't have "taste", that is, they aren't good at producing "minimal" descriptions of behavior that we have historically found to be resilient predictors for new information. (Scare quotes mostly because actually measuring this kind of thing is a nightmare) Reply @brabra2725 3 years ago Pattern has two Ts. Reply @MountainFisher 3 years ago A DNA molecule is 2 meters long and when folded the genes are next to the genes they code with. Some in the huge ENCODE Project discovered that the coding genes work with whatever genes they end up next too. Watching the interactomes working showed that there really isn't any Junk DNA. It all has some function, we just don't know what it all is...yet. 1 Reply 3 replies @SteveBakerIsHere 2 years ago AI is like magic - but it's not ACTUAL magic, it's guaranteed to make mistakes. The real question is whether those mistakes overwhelm the benefits when it does get it right. I'd hope that it gets good enough to earn money that would pay for more data - because data is the driving force behind AI. This kind of strategy is what got Tesla's AI car driving software to work. You ship something that's just good enough to be useful - then use that to collect data to improve the AI. The better it gets, the more data can be collected on the cases where it gets it wrong - and the faster it learns. If they made a piece of software that could make fake electron microscopy results from the folded protein predicted by the AI - then electron microscopy can be used to capture simple images of the real thing that can be compared to simulated microscopy of the predicted result. All the AI needs for training is "You got it right!" or "You got it wrong." - you don't even need a human to "interpret" the microscopy...so data could be obtained from any sequenced protein with just a photograph. Reply @AuaerAuraer 1 year ago Nice Reply @toddbellows5282 3 years ago To bad that Google is so evil. They had such great potential. Reply @elimgarak3597 3 years ago (edited) 10:02 Well, researchers also need data to build models. I don't think that is a valid criticism. Reply @GC-kw1gq 3 years ago What is the probability of Protein Folding happening by chance in a universe happening by chance? Lambda squared? Lambda raised to the power of lambda? Oops Theory Reply @sobreaver 3 years ago There is potential, in everything, to be 'good' or 'bad', whether it be an elected president, a kitchen knife or a chosen AI, it is up to all of us to know how far we let things go. But none of these are actually bad in itself, it is just the way it is used. AI can be quite useful, but we wouldn't want to blindly rely on it more than anything else in the world. Results must be tested again and again, and guided toward OUR Needs (and hopefully not Our GREED oo ) Reply @sanjosemike3137 3 years ago I just heard Dr. Hossenfelder say that “artificial intelligence cannot create a new function without an exterior direction for that change.” Maybe I misunderstood her. But if she really said that, how has the protein folding problem been solved? Great video though! Sanjosemike (no longer in CA) Reply 1 reply @georgysb 3 years ago What a bright future, all the biologists will soon serve as a measurement units for AI researchers.=D Reply @nilsbeckmann3608 2 years ago As the goal is 100% correctness the 90% could very well direct you in a wrong way. As long as there is no real A.I. which can emulate human thinking I don't think they help solve such abstract problems where you would actually search for an overall theory. Even a 99.9% correctness might be a randomly close result to a theory which must be formulated in such a different way, that you won't get nearer to it just by having ways to calculate an example very, very close to their prediction. I would like to understanding protein if folding is important in research for "real" A.I.; i.e. simulating neuronal networks at the molecular(or atomic?) level . I think that was the path to go to solve real probable A.I. is/was going (with predictions it might be solved ~2025) I saw in a video ~5-10years ago (or maybe read somewhere). It would be funny if progress in each of both would require progress in the other. Reply @jordansmith3809 3 years ago holy shit i fucking love this Reply @realcygnus 3 years ago Cool Channel. Reply @catmate8358 3 years ago (edited) Sabine: use two horizontal lines and two vertical lines to split the screen into thirds. Place your head at the intersection of the upper horizontal line and one of the vertical lines and stay there. Use the free side of the screen for illustrations. You need to stop jumping left and right across the screen. Whoever advised you to do that is WRONG!!! Trust me. I'm a filmmaker, I know what I'm talking about and I'm here to help. Your content is valuable and unnecessary visual and editing artifices only distract the viewer. It's ok for politics, but not for science. 2 Reply 2 replies @Richard-vw8no 3 years ago I wonder if a book on origami will help me understand this topic? Reply @michael.forkert 3 years ago And if it was? What the difference would it make for the million children which die from malaria in Africa per year? Reply @deanerhar 3 years ago I tend to fold my proteins in several layers of tissue paper. Reply @sols4102 3 years ago could it be possible to build a computer made of protein Reply @jacquesfrancois4275 3 years ago 3:32 Haha oooo shots fired across the bow! Reply @ElElwe 3 years ago Awesome Reply @user-or6oo2hm9r 3 years ago Wie geil! Reply @Dablid77 3 years ago You are perfect Reply @jaredbaum 3 years ago Have they tried a Middle Out approach? (only Silicon Valley fans will get this) Reply @AndreasHLux 3 years ago Professor_in auf YouTube Stufe. Reply @posthocprior 3 years ago She didn’t answer the question if the protein folding problem has been solved. Reply @joehinojosa24 3 years ago SO IT TAKES ARTIFICIAL INTELLIGENCE TO UNDERSTAND NONINTELLIGENT CONSTRUCT? Reply @LivinTheDreamOfDan 3 years ago If a bad fold can cause ALS and the COVID vaccines create new proteins could the vaccines cause ALS eventually? Reply @babyl-on9761 3 years ago Are these protein chains living or are they inert pre life molecules? Reply 4 replies @JiveDadson 3 years ago "Solved" is definitely the wrong word. AlphaGo can beat any human at Go, but it certainly has not solved the Go problem. Updates to the training appear to make incremental progress. Reply @28DAYS77 3 years ago Curse of the scientist that last 10% does make the difference! I do remember the AI video you did you looked beautiful and it was a great video your a amaing scientist Reply @davizitopa7252 3 years ago I think they need to perfect ab initio chemistry (not sure if this is the correct spelling). It takes a lot of computation and is currently very inaccurate. Now that I think of it, its inaccuracies probably come from the inherent probabilistic nature of quantum mechanics, maybe? Hard not to think of quantum mechanics while looking at Dr. Hossenfelder's face If I needed an impromptu excuse right now it would probably come out as "quantum mechanics ate my homework" Reply 2 replies @edcunion 3 years ago Everybody is a local island universe of actions sandwiched halfway between the quark-gluon and CMB thresholds! Glad the proteins fold on their own initiative while being battered by Brownian effects without our having to consciously think about it! Proteins are clever and hard-working little soft machines, no? So was David Bowie crooning about genetic code when he so eloquently sang about the gene genie strung out on lasers screaming and bawling and letting go? It's still mysterious, what are the chances DNA, RNA, amino acids and proteins could figure out what they do in ~13.8 ga through random actions and events they had no control over? Since we are here discussing this it seems 100% at least for the time being? Still thunderstruck here being a conscious organism that played no part in its existence except perhaps being the unmeasurable desire that brought two similar thinking organisms together that were a good fit? Adios for today amigos, thanks to the thought provoking Dr. for sharing weekly! Reply @chefbennyj 3 years ago So all those computers running "folding at home" last year helped with this science? Is that how they did it? Reply 1 reply @iamable915 3 years ago Alphazero went gaga on chess Reply @sakubashiba3610 3 years ago uv klights make them follddd duuhhh.....like the solar radioation or heat... its stress and tension.... i mean its in the name folding like metall...same stuff...sun folds them or heat fods them ...get it ?soooolved..also they twist they dont fold... Reply @justmemyselfandi7760 2 years ago What I don't understand is how the theory of evolution on the basis of natural selection can explain any of this within the timescale of life on earth. I mean, a single biological entity has tens of thousands of proteins with very complex interactions between themselves. The probability of generating an amino acid sequence that creates a single correct functional protein through natural sampling seems absurdly small. Sure, given an infinite timespan a miniscule probability could continue to exist... But considering it's only been a few billion years since the earth even came to being... This does not intuitively make sense to me. Reply @TheHiddenMarty 3 years ago Could we train AI to create the theory of everything? Reply @michelandre8106 3 years ago 1 Reply @jjeherrera 3 years ago 10^300? That's nothing compared to the 10^500 possible M theories... :-) Reply @DAVHORNER 3 years ago This sounds like a problem for quantum computing to solve. Reply @janerussell3472 3 years ago We can no longer make vitamin C. That implies our diets were fruit rich in the past. We've lost the capacity because it wasn't essential, perhaps. Other animals that can't make vit C are heavy fruit or green grass eaters, like fruit bats. Lactose intolerance is another loss of function. Adults from Africa and Asia may be intolerant to dairy products. Does that imply Westerners were weened longer? Certainly I was still sucking my dummy until I went to school. The last Emperor of China wasn't intolerant. According to the film, he still had a wet nurse as a teenager. The rest of us can only dream. It's thought the change to meat eating, as we swung down from the trees onto grasslands, was instrumental in the exponetial growth of the human brain. Lucy, clearly an australopithecus, had a brain capacity of about 450cc. We have a skull capacity of maybe 1400 cc, Neanderthals slightly larger. Einstein' brain was 1230 grams, at the low end of average for modern humans. It's not what you've got but how you use it, apparently. My brain is my greatest erogenous zone...unfortunately. No one can see it. lol. That's all. Reply @vladyslavkorenyak872 3 years ago Could the protein folding problem be done by quantum computers? Could we program "quantum neural networks"? Why do I have more questions than answers? Please stop my braiiin! Reply @YouMockMe 3 years ago Wow...wait. So the fold of a protein (the "same" protein) can mean the difference between Alzheimer's or diabetes??? Reply @harrymason1053 3 years ago Great video but it's pace is too fast. Reply @tobiasthrien1 3 years ago 5:25 Isn't that mixed up!? A bottom-up approach would use an underlying model from which you can work your way "up", while a top-down approach would start by observing the end result at the "top" working your way "down". Reply 6 replies @catStone92 2 years ago not all artificial intelligence makes use of machine learning Reply @gazzaka 2 years ago (edited) And what's the half life of a color tv? Order a mexican meal, it's all basically the same, just folded differently ! I am working on Artificial Stupidity Reply @andrewroberthook3310 3 years ago Sounds like the macro version HUMAN BEING Reply @Bassillixx 3 years ago my wife unfolds my protein at least once a week . . . . 1 Reply 2 replies @zahiddogan 3 years ago Are we trying to learn how proteins fold to make new useful proteins that never seen ? I didnt understand where to use this ability. Reply 1 reply @squamish4244 1 year ago (edited) Since this video was made, we've made even more advances, and more are imminent. Good lord. It reminds me of Louis CK on how people become instantly frustrated with things. Would you give AI a second? Like, give it a second? Update: AlphaFold 3, which still hasn't 'solved' protein folding, but is significantly better than AlphaFold 2, was released three entire years after AlphaFold 2. Reply @kalaimuthu 3 years ago Great subject matter but my eyes are tired ... and now I am annoyed Reply @ericsbuds 3 years ago basically the chance of a protein folding on its own is the deciding factor of life arising on a planet xD jk Reply @mrgyani 3 years ago (edited) The things AI can do now is absolutely chilling.. The implications are huge, almost mind blowing. And to realize all this power is now held by one company - Google Reply @liammurphy2725 3 years ago My wife laughs when she sees me folding shirts. Can this research help me? Reply @bhspenceryt 3 years ago Type 2 diabetes, how so? Reply @dr.merlot1532 3 years ago it still hasn't helped scientists cure ALS though or any other diseases Reply @Drifter4ever 1 year ago Very funny that they point out the ridiculous random possibilities and have to put in tons of AI for prediction of usefulness. Biologists still don't understand they will step by step prove the one thing in life they hate and fight so much; intelligent design. Reply 1 reply @hschultz123 2 years ago Not to take away from this video, but. People do not need to eat any protein. The body breaks down proteins into amino acids and then reassembles them into various human proteins according to the DNA code. The human body can manufacture all but four essential amino acids, which must be eaten. 1 Reply @mujtabahassan4090 3 years ago Succinct and clear Reply @Tymeshifter 3 years ago (edited) 7:42 Why the abbreviation A.I does not have a period after the I? 2 Reply Sabine Hossenfelder · 1 reply @u2mister17 1 year ago You didn't explain how a "protein" is formed. Stating from the middle of the question is doesn't end with science. Reply @nobiggeridiot 3 years ago Isn't utilizing AI in these types of scenarios, just the pragmatic 'bypass' ? Does having an AI that can reasonably predict how a string of proteins might fold, tell us anything about why the string of proteins folds the way it does ? It sort of feels like cheating on a test. Or is it that we already know all the answers to the test, but that it is so incredibly long that we need a computer to do the work bits ? 1 Reply 1 reply @FighterFred 3 years ago AI is a black box with no predictive power and you cannot follow what goes on inside the box, so the result cannot be explained. As for new drugs, I'd rather build a molecule step by step by adding atom after atom and test its biological impact in every step. This can also be done through software. Reply @manonthedollar 3 years ago Sabine u so stylish Reply @nathanielmcdonald1910 3 years ago so we gotta wait for quantum cpus? once it's figured out... is it really that simple to fix Alzheimer's and diabetes? Reply @mntrmntr 3 years ago love the accent Reply @Krmpfpks 3 years ago If you're only here for the string theory burn: 3:32 ... you're welcome. Reply @markh.harris9271 2 years ago (edited) Yes, AI is only useful when the neural net has enough data; but its worse than that... the programmers who tweak the net need to "already" understand the problem 'better' than the deep learning system in the first place, and we're no where near that, unfortunately. There are more than one AI technique; neural net is just one. (just saying) Discussing protein folding without discussing ribosomes is ridiculous in my opinion; that and the coding that controls the ribosomes needs to be included in the discussion. Amino acids will not fold properly without ribonsomes... um... you can't properly fold amino acids in a test tube! The process of folding amino acids in ribosomes is extremely complex machinery which unfortunately is also not fully understood. Amino acids have left and right handed versions of themselves; only the folding type is synthesized in the body, and only the not folding type occurs in a test tube (or in a prime soup, for that matter). When the folding malfunctions it can be the chain, the amino acid (individual) , the ribosome, or the coding; and that is also not fully understood (to say the least). AI will only be able to help in this problem area "after" the knowledge base grows some considerable amount . (?) marcus 1 Reply @MeppyMan 3 years ago 3:32 is this a string theory dig/joke that I’m just not smart/educated enough to get?? :) Reply @eug_metta 3 years ago Great segment but "NordVPN keeps you safe as you browse the internet" is about as true as "string theory has solved the black hole information paradox".... Actually keeping safe on the internet involves lots of rather complex factors and while NordVPN seems to provide some degree of protection there are still many ways one can get hacked even when using NordVPN. For example through sites that aren't on their blacklist, infected sites on Tor, malware apps and dodgy email attachments. Reply 1 reply @droneready8278 3 years ago She has some beautiful hands Reply @michaelhusar3668 3 years ago I Know so little about biology, that I can safely say I didn't understand this video, but I will be careful not to fold my peanut butter sandwich so I don't mess up the protein Reply @yrebrac 3 years ago (edited) Very lucid Reply @ogvibe11 3 years ago I wish you were my teacher Reply @Merilix2 3 years ago What is the scientific value of an AI-answer where no one knows or can recognize how that AI got that answer? 1 Reply 1 reply @Y2Kvids 3 years ago Post Scarcity society Reply @Pod-042 3 years ago In the future, skynet will "fold" alot protein. Reply 1 reply @alcoholandfun243 3 years ago And somehow it all happened naturally in a big bowl of soup ️ 1 Reply 1 reply @richlinlaw 3 years ago accent is like Meena from Veep ie smart/sexy Reply @kingrobert1st 3 years ago Has protein folding been solved? Obviously not or the header would have been Protein Bending Has Been Solved. Big difference. Reply @beardedroofer 3 years ago Ok, great vid, but I'm lost. What is replicating folding proteins good for? How would that improve life for the average human bean, or solve hunger, or eradicate cancer? 1 Reply Sabine Hossenfelder · 12 replies @Salvaeshun 3 years ago Sweet hoody! Reply @europaeuropa3673 3 years ago I want a protein that gives me immunity to the pneumonia bacteria for life. 1 Reply 3 replies @gregsansenbach705 2 years ago MORE Reply @darrennew8211 3 years ago I want to know of you, Doctor Hossenfelder, get to keep all the outfits you show off, or whether you just rent them for the filming, like at the Oscars? :-) Reply @thedecktothe16thpower56 3 years ago (edited) Plutonic shapes may dictate how a proteins crystalize or how they are prepped at a super position state for maximized exchange depending on the patient and cells. If one was crafty you could vacuum forge a protein chain trying to freeze time as much as possible allowing natures natural vortices of hidden time in the future to say hello can I help you. Yes, that will be one hat trick please with no cheese or mayo and hold the soft drink. Nothing like pure mana from the tap to revive ones sprites and shuck impostor cells that are running away with your time. Perhaps with the right frequency the bad cells could be walked right out of the flesh with treatments. At the same time tuning into the marrow of our bones to make up the indifference of the bodies pressure on a plank scale through blood. Of course this has to be pretty groovy for a heart frequency. As a matter of fact thats the perfect spot to base the process and work out from there. Drink lots of water you'll be thirsty. Reply 6 replies @scottfranco1962 3 years ago After we solve the folding problem, we need to move to the spindle and mutilate problems... Reply 2 replies @martinda7446 3 years ago Sabine, will you go for a drink with me? I always fold protein when I put ham in my sandwich. Just saying. (Sorry everyone for being such a twit, I can't seem to help myself. Any constructive suggestions to remedy this personality disorder are welcome, including jumping in the lake etc.). Reply @MrAlanCristhian 3 years ago 3:32 brutal 2 Reply @elbertsplee7937 3 years ago AI is the new oracle? Reply @janerussell3472 3 years ago No one is interested in my 2-pence worth, I'm sure. 2 points, though. 1/ Denatured proteins still know how to re-fold, it seems. 2/ Results obtained by machine learning techniques (with many adjustable parameters) on small training sets are very good, but results obtained by these techniques on testing sets are very bad. When the training sets grow, results obtained on them become worse, No doubt AlphaFold is slow compared to rival techniques. [ AlQuraishi’s system, which uses an algorithm called a recurrent geometrical network (RGN), can find protein structures a million times faster—returning results in seconds rather than days. ] But it's more accurate. No doubt if they incorporate the Fisher sampling equation it will get faster. If protein folding can be predicted thus, one can't help thinking that folding- to be so quick in real biological systems- is pre-programmed. Just a thought. The oragami master, to fold a butterfly, already knows the folds to make. As I say, just speculation. [ fits in with an intelligent designer, though ] Reply @vperez4796 2 years ago (edited) I am not going to vote this time. Reason No 5, X-ray diffraction data comes from "crystalline packing". There aren't protein crystals inside a cell. No 4, Using NMR data in solution, you find that there isn't a unique conformation of the native protein. No 3, protein extracted from living cell are commonly stabilized with surrounding companions, they don't exist in a COMPUTER VACUUM. No 2, Gibbs free energy predict that there are several wells before you hit the lowest point in energy. Please ask how many folding conformations have "PRIONS"?. No1, Pharmaceutical companies can predict reasonably well a structure, BUT, they cannot predict if a medicine is going to fit in THE target protein, neither which how stable is the protein-medicine. In Pharmaceutical science, they can't tell if albumin is also going to bind a medicine, nor how tight is it going to fit in. I am sure that IA can find which model is most accepted by Pharm Companies, before the clinical trials. Reason No 0, researchers have always a way to explain slow progress they achieved, so they publish preliminary results. Reply @pauljasmine353 3 years ago We can put a man on the moon but we can't figure out how proteins fold. Reply 1 reply @michael.forkert 1 year ago Frau Doktor PhD in EveryThing- & Teleprompterlogy rides again. Reply @ClitaLarue 3 years ago It was solved ten years ago Hortensia Reply @ThomasLee123 3 years ago AlphaFold AKA Google. Ugh! Reply @oldranger649 3 years ago deep learning uses neural networks. what the hell does that mean? Earlier Video I guess. Reply @l27tester 3 years ago Put the benefit of the current state of AI protections narrow down the set that the human experimenters might use to further refine the model? Almost to too fased approach Reply @johnnytoobad7785 3 years ago The first "intelligent" life form was just an RNA molecule that eventually just "folded" into humans ? It just took a while. (in non-relativistic time...) Reply 2 replies @miguelpereira9859 3 years ago Machine Learning will change EVERYTHING Reply @colfaxschuyler3675 3 years ago ?? Protein origami? Reply @davidwvalentine8024 3 years ago AI Isn’t really AI now is it.? Reply @eugen10min 3 years ago who's Nobel then? Reply @peerajak 3 years ago Can we create a new type of animal from this protein folding? Like for real Reply 1 reply @ultravidz 3 years ago Headlines are always misleading and should always end with a massive asterisk Reply @paulmumford3397 3 years ago two t's in pattern Reply @bobcabot 3 years ago ...nice Pulli! or jumper to use the right language,,, ( wanna get noble... Reply @TheNameOfJesus 3 years ago Sabine is generally good, but @6:17 what's a "deep frozen protein"? And how does that differ from a "frozen protein"? How can a single "molecule" be frozen? (She shows a single molecule at 6:22, which illustrates that she was talking about a single molecule.) For comparison, a single molecule of water, by itself, can't be frozen. The state of being "frozen" is simply what happens when MULTIPLE molecules join together. Is this what she means by "crystallized proteins" @6:08? I like Sabine, but there are so many verbal ambiguities in this video I cannot say this video is up to her normal standards of clarity. Reply 1 reply @daibangx8696 2 years ago Well, today DeepMind, in collaboration with others, just announced that they "solved" 3D structures of 200 Millions proteins. Before 2021, the database is only nearly 1 million. One year later, its 200 millions. AI is no doubt will be the savior of mankind. But wait, who created AI? Isn't that mankind? Reply @humanbass 3 years ago We just want better PEDs, moar plates, moar dates. Reply @kristfallon9989 3 years ago So protein shakes are worthless? O.K., got it. Reply @sapientum8 3 years ago Now it's no longer "God of the gaps" - these days, it's "natural selection of the gaps". The universal answer to any unknown biology question is, "obviously, - it's because of evolution and natural selection." Even when this clearly makes no sense... Reply @JackSparrow-nq5wh 2 years ago Covidvax causing prion Reply @kingkhann9 3 years ago When stupid youtubers make some video to make money they make stupid videos... this is an excellent example of that... Reply @weerobot 3 years ago Protein Fold a Big Mac..lol Reply @brainretardant 3 years ago There are chaperones that are induced under different environmental triggers. Glycosolation rates and selection vary as way. This is a pipe dream. Enjoy your grant money and fudge your results to keep it coming Reply @serbrat 3 years ago ...or A.I. will help develop new designer drugs much faster Reply @jamesr2408 3 years ago The duplicity of atheists know no bounds ... O Lord my God, When I in awesome wonder Consider all the worlds Thy Hands have made I see the stars, I hear the rolling thunder Thy power through out the universe displayed Then sings my soul, My Savior God, to Thee How great Thou art, How great Thou art Then sings my soul, My Savior God, to Thee How great Thou art, How great Thou art Reply @terminatedaccount8750 3 years ago Can it fold the protein spikes on the 501 v2 variant? 1 Reply @sylvainbouchard2115 3 years ago Tea. Earl grey. Hot. Reply @toughstuff3374 3 years ago Bad explanation of the vpn. Very misleading. Reply @mirimirmiri3612 3 years ago Spiders web. What? Web. And? Spider. So? Proteins. Explain. Folding. Am. in. o. ac. ids. Reply @monad_tcp 3 years ago (edited) No, protein folding was not solved, its a NP-complete problem, if you solve that, you solve all NP complete problems too. Its heuristics, didn't solve the problem, kind of skived it and came to something closer to the real solution, but with an error delta. Reply @Mechulus 3 years ago (edited) I use the bottom up approach... ... WITH MY GIRLFRIEND!~ Reply @rodnorris9532 3 years ago I am glad to hear you break the myth that protein is just meat to eat! 1 Reply @noircc 3 years ago Net schlecht, sehr interessant Reply @phils744 3 years ago Hello everyone, I'm "No Expert myself" (I do enjoy listening to this podcast, %100). Please. But JJ pickle out of Austin, being a "Super computer" (being massively parallel) ran the complete human genome in under 30 minutes, but (very weird, the PS3, networked game console, the fastest network on the planet doing how many Petaflops ) can do real time gene folding just saying. Be safe everyone. Reply @Kwitzats 3 years ago tldr meh it is just one step in the process Reply @alloomis1635 3 years ago oh, dear! driven by hunger, or insatiable narcissism, dr. h has left the relatively clear waters of physics and dived into the murky fluids of biochemistry. a field in which human understanding is hopelessly incompetent to encompass. perhaps it is a tentative experiment, to generate funds for a mediterranean holiday? Reply @RP-mm9ie 3 years ago Avi Loeb Reply @kldogol9053 3 years ago I don't know why this video recommended to me, but... why are you look like The Queen? cool video tho Reply @dianehansen5552 3 years ago Trying to work out the accent. What nationality would this woman be? Reply @macdougdoug 3 years ago Could we have a punchline rimshot for any scientific jokes? Either that or better comedic timing - I feel I'm missing out on some of the in humour. Reply @gyro5d 3 years ago Amino acids are magnetic. "The Ebner Effect" = Evolution. Reply 1 reply @japexican007 3 years ago A dark winter indeed Final warning! This is it!!! 1. Scared of how you’re going to pay your bills? 2. Afraid of the impending economic collapse? 3. Feeling distraught due to the inevitable breakdown of society and eventual civil war? All these things happening now are but the beginning of sorrows and this is Gods way of giving you a chance to be saved before time runs out!!! “Behold, I stand at the door, and knock: if any man hear my voice, and open the door, I will come in to him, and will sup with him, and he with me.”‭‭Revelation‬ ‭3:20‬ ‭KJV‬‬ Things won’t get better, this is an illusion the powers that be have placed upon the world, once the rapture of the church occurs it will get vastly worse, but it’s not too late for you to be saved even now, all you have to do is: As a repentant sinner Look into your heart and confess: Dear God I am a wretched sinner, as I repent I put my faith that Jesus is God and that he died buried and resurrected so that his blood can wash away my sins I put my faith in that alone to save me not my good works in Jesus name I pray Amen. Romans 10:9-10 "That if you confess with your mouth, "Jesus is Lord," and believe in your heart that God raised him from the dead, you will be saved. For it is with your heart that you believe and are justified, and it is with your mouth that you confess and are saved." Don’t put your trust in this world, put your trust in God “Heaven and earth shall pass away, but my words shall not pass away.”‭‭Matthew‬ ‭24:35‬ ‭KJV‬‬ Reply @eulefranz944 3 years ago 3:31 The perfect women doesn't exi- Reply @BrianThomas 3 years ago Can these new mRNA vaccines cause proteins to become folded over time? Reply @KristofferEngstrom 3 years ago Im all in for AI, as long as we before this tech is widespread, has access to AP rifles and emp grenades ;) Reply 1 reply @benquinney2 3 years ago Turbulence Reply @lukapopovic6668 3 years ago Could a quantum computer resolve this problem? Thanks Reply @040_faraz9 3 years ago Hope smart people think more about contributing to major problems rather than working for multi billion industries. Reply @divvy1400yam600 3 years ago Being an expert on almost nothing am I wrong in saying that when proteins produce the desired results to enable say the correct operational structure of the human body then the linear polypeptides produce by RNA/DNA chemistry will have changed thru 3 or 4 levels of folding. Then the folded proteins have to connect 3 dimensionally with other folded proteins in order to 'work' This was my limited understanding of Hoyles work I spoke about earlier. All this is supposed to have occured 'naturally' which leads me to conclude that most who so believe lack imagination or the temerity to say what is true ! Natural Selection eat your heart out ! Because clearly I did not know this level of detail quote: Alzheimer's disease is an example of a neurodegenerative condition caused by protein misfolding. This disease is characterized by dense plaques in the brain caused by misfolding of the secondary β-sheets of the fibrillar β-amyloid proteins present in brain matter. Huntington's disease and Parkinson's disease are other examples of neurodegenerative diseases associated with protein misfolding Reply 5 replies @miguelalfonsomartinezbarra421 3 years ago Fan... :3 Reply @charleswettish8701 3 years ago So.....she's saying we'll be able to eat plastic, soon??? :) ;) ;) Reply @rwbyrose7269 3 years ago I still just watch this because you are cute, I can’t keep up with the science Reply @victotronics 3 years ago Did I spot a quick dig at string theorists? Reply @mariostelzner4530 3 years ago YES PROTEIN FOLDING HAS BEEN SOLVED. BY CELLS. AHAHAHA AHAHAHA LOL Reply @bobhoven3959 2 years ago Your body tells you , some nutrions work better then others !@@ also with milk and eggs original is better Reply @nomorestealing7152 3 years ago Let's talk about why humans have only 46 chromosomes Reply @firstcommenter202 3 years ago AI has not solved the laundry folding problem Reply @janerussell3472 3 years ago We can unfold proteins-- and remember, the likes of Dennett, Churchland, Pinker, Dawkins, etc,. think you are just meat and bone-- but we don't know how to unfold our love. Reply 1 reply @Theineluctable_SOME_CANT 1 year ago I was hoping for a quick yes or no, babe... Reply @vedantsrivastava5754 3 years ago You should do movie criticism not science Reply @gozer825 3 years ago Am I a nerd for finding this interesting? Reply @tiglet53 3 years ago chic Reply @terrywbreedlove 3 years ago Isn’t this how the Covid 19 Vaccine was developed. Reply @thygrrr 3 years ago (edited) Hmm, I think even you are perhaps not immune to the typical human fallacy of orders of magnitude looking similar in exponential notation. 10^300 protein folds are nowhere "almost" as many as 10^500 vacua, they're not even nowhere almost as many as 10^301. Just as 1 egg is not almost as many as 10. Reply @janerussell3472 3 years ago DISTINGUISHING BETWEEN FACT AND TRUTH Even when we see straight we can't always believe our eyes. We only catch light at little more than 25 frames a second when we see things as continuous. But that's not what I mean. Nor that our brain turns input the other way up i.e. that it's processed. What I mean is that, for example, we see the Sun rise in the East and set in the West. But that's not how modern people interpret it, since Copernicus. tHAT'S hALL, y'all. I'd love to stay But I have promises to keep, And miles to go before I sleep, And miles to go before I sleep. Reply @ThomasConover 1 year ago The “kernel” of God’s protein-based machine-code is a mindblowing impressive and well kept secret. God is the greatest engineer of all. Reply @Tripilen 3 years ago 666= goats gold girls 369=love Reply @Nulley0 3 years ago Nobel prize winning discoveries and inventions will be made by an AI called "TAS" Reply 3 replies @mikaelsjoberg1894 3 years ago (edited) So now Google becomes God - for real. If that not is the most frightening thought you ever had i dont know what is. Reply @michaelqiu9722 3 years ago This video comes across as underwhelming, perhaps because Sabine herself doesn’t do biochemistry. You should have at least touch on the fact that there are 20 common amino acids, each having different properties including size, polar/nonpolar, charge and so on which interact physically to decide the low energy conformation (native state) of a protein. Should have at least explain what’s the primary sequence of a protein. Blumenthal’s paradox has little to do with the difficulty of solving protein folding, because that’s not how it’s approached computationally anyway. Reply @satzumichaeng8474 3 years ago (edited) Maybe Artificial Intelligence can help answer a couple of recent and seemingly intractable questions. How was the novel corona virus created & released, and, who received the most “legal” votes in the recent American presidential election? 1 Reply @VideoNOLA 3 years ago NordVPN may be a popular channel supporter, and that is laudable, but for anyone watching, it's completely unnecessary! Reply @Jeshua1737 3 years ago Underlines why such complexity can only be the product of Intelligent Design Reply 21 replies @joshcarr2356 3 years ago I wonder how many people got your severe string theory burn Reply @deldia 3 years ago She doesn’t say Einstein in this video Reply @lulo08 3 years ago You look beautiful. Reply @datinsky69 3 years ago So, let me see if I understand what the video is saying. Humans already know a set of solutions for the protein folding problem and A.I. guessed 90% of what we already know? Reply 1 reply @s9209122222 3 years ago Answer: No. Reply @airman122469 3 years ago Answer: no, not even close. Reply @ochsblogger 3 years ago Couldn't physics be helpful? Certain permutations of the folds could be dictated by charge, covalence and other electro-chemical properties of the adjacent amino acids reacting with each other? Reply @tTtt-ho3tq 3 years ago One question to any biologist. Protein folding is still part of evolution, right. So there's no intent. What ever the proteins there're today survived through have been evolved by natural selection, right? Reply 1 reply @su-1337 3 years ago Why YouTube recommend this? Reply @clmasse 3 years ago Anyway, protein folding won't solve the problem of life, it is still way more complicated. Reply 1 reply @gaiustesla9324 1 year ago HEY! REACHERS! LEAVE THAT GENOME ALONE. Reply @hank1475 3 years ago (edited) What's a happy tool??? You are it. Where would existence be without life...dull. Reply @r123brown 3 years ago I’m much smarter now. Reply @harveypost7799 3 years ago Y u do ur own laundry... Reply @fredsmith2277 2 years ago she's wrong about something..... i just dont know how yet, but i vow to dedicate the rest of my long live to proving her wrong, i better not, that would shatter her ego and cause great distress for certain !!! Reply @davidnaugler73 3 years ago Solved by proteins. Reply @kennethbransford820 3 years ago How was folding protein even possible without design? ==== Evolution = Self Assembling Atoms = Impossible ===== 1 Reply 16 replies @redfishervictoria 3 years ago Understanding Magnetic geometry is the key. Reply @Mike-hr6jz 3 years ago It’s not been solved no matter what the bull crap you with Reply @acemanNL 3 years ago Cat X Cat X Cat X Tortoise V Reply @canemcave 3 years ago how many string theories are there? Reply @workin4alivin585 3 years ago (edited) Damn monopolistic Google. Tentacles in literally EVERYTHING. 1 Reply @DefaultDerrick 3 years ago (edited) I love the german people. So brilliant, so strong. Germany is the shining jewel of europe. So long as they avoid the mistakes of the past, i.e. nazism, military empire, etc. Nothing can stop Germany or dim it's influence in the world. Look how brilliant this german woman is. And that accent! Reply @jamilfromyemen9886 3 years ago why do not translate your vid to arbic Reply @jamilfromyemen9886 3 years ago why do not translate your vid to arbic Reply @marshallodom1388 3 years ago No, and I'm not going to bother watching. Reply @janerussell3472 3 years ago (edited) Reductionists think we are just "a bright bracelet of hair about the bone"- meat and bone in other words. Leonard Susskind goes further and thinks we are holograms. Others, like Nick Bostrom, think we are SIMS. We're not SIMS, or every cell wouldn't have the DNA code. I'm reminded of the film 'Toys' where the kids can't figure fantasy from reality, and have no moral qualms about using war drones on real people. It's just an arcade game for these nutters; and we're the target. I include religious nutters who think their God is telling them to murder. God does a lot of killing in the Bible...but a Master Potter doesn't usually keep cracked pots. But let's not get into that. Reply @oddarneroll 3 years ago Informing about science segwaying to misleading about security. If it can't be done in a honest way, it's not worth it. You should not give knowledge with one hand, missleading with the other. Reply @jaydawg7820 3 years ago Wow, almost a Centillion folds ... Origami has nothing on Protein ... just when you thought the Universe was complex you ad 300 zero's ...... Reply @dimitriosdesmos4699 3 years ago a change from a glutamine to an aspartic acid can result in total disaster in terms of protein functionality therefore anything less than 100% is a total failure. Also in eukariotic cells proteins need a scafold to fold and they fold in different cell compartements with different environments. From a fundamental knowledge perspective this is brilliant though,...anything above 60% consistently is awesome.....I still think that reverse engineering is more practical...infinitely more practical than trying to predict...For prediction, there is an unholy approach which i am not going to mention on fucking youtube enit ? Reply 1 reply @WaltersGuides 3 years ago is she an AI? Reply @Nosferatu186 3 years ago Protein folding should be left to the realm of God. We will never figure it out, God is beyond science. Reply @orangesite7625 2 years ago First time I read it PROTON FOLDING 1 Reply @cousinbryan3007 3 years ago In short, this is the argument from design which proves evolution fails the test of science. Maybe it's time to give Genesis another look. Reply @John-qx1xn 2 years ago LOL Reply @jesus-ck4el 2 years ago Why not make the AI predict the question instead of the answer Reply 1 reply @Fisherofmen64 3 years ago DNA proves the existence of God. DNA is information. Information is facts gather about something. Facts are made up of words. Words are thoughts expressed. Words are spiritual and can only be created in the mind. You cannot speak or write a word unless you first create it in the mind. DNA are words created by the Mind of God. God is the Programmer, DNA is the software instructions contained in seeds that instruct it on what and how to become. Reply @alext5497 3 years ago I cannot wait to be used as a genuine pig by the government. Reply @RalfStephan 2 years ago (edited) Unfortunately, incorrect protein folding is the cause of a only small subset of Alzheimer's disease and ALS. The real cause of both diseases are still a matter of research. And no, the amyloid hypothesis is not satisfactory. Reply @MrofficialC 3 years ago Just like with the number pi I'm willing to bet there are segments that repeat and will bend in the same way Reply @adrianflo6481 2 years ago wow, a VPN ad without any fear mongering. Strange sight to behold. Still worthless tho Reply @arturodelosangeles 3 years ago so sexy intelligent beautiful woman , with such a beautiful accent Reply @robsmura888 3 years ago OMG, stop moving her all over the screen. Just hold her in one place and use the other space to display other graphics. Holy headache. Reply @miketierney7510 11 months ago Not until energetic, genetic and molecular ideologies merge into one comprehensive understanding of how an organism as complex as a human being can be deliberately designed constructed and made autonomous! Then we will have solved the problem of protein folding. But we still will have not solved the mystery of who was it's original creator...until we find God. Reply 3 replies @ezrasteinberg2016 3 years ago And yet more proof that quantum mechanics allow living things to accomplish critical survival tasks: https://www.youtube.com/watch?v=ia5vxCSiwm4 Reply @ambrosechiel6931 3 years ago This what the Sovereign, Supreme; Absolute, All-Being, Almighty Power The Most High GOD of Yasharala says :Tell this to your Scientists, With my Great Power and outstretched hand(ArcAngels), I Made The Earth, and it’s People and Animals that are on it, and I give it to Anyone I please. For my thoughts are not your thoughts , neither are your ways my ways, declares The Almighty Power The Most High GodYAHAWAH, as the Heavens are Higher than the earth, so are my Ways Higher than ways, and my Thoughts than your mortal thoughts. Dr. Ambrose Chiel, PhD. 1 Reply @10spen11 3 years ago "Huge if true" Reply @jamesr2408 3 years ago @4:28 very interesting subject, but I gave up after she mentioned natural selection. Reply 5 replies @theobolt250 3 years ago Very exiting stuff in more then one way. It looks like it that solving the proteïn folding problem 100 percent cannot be achieved without the next best thing in AI. The consequences of this may be more far reaching then one would expect. Reply @hoaxuan7074 3 years ago (edited) I fold them by boiling an egg, dear. Reply 2 replies @Mayadude66 3 years ago Interesting, but you are making a common mistake by using the word ''function'' incorrectly. Function presupposes a designer. Thus, proteins designed by scientists have a function. Proteins in our body do not have functions, as they were produced by evolution. By saying that proteins (or organs for that matter) have functions, you are saying that they were made for a specific purpose (a teleological view of nature) and so you are speaking as a creationist. A simple example makes it clear: A river can create a lake, but that is not it’s purpose. A canal can create a reservoir, and that would be it’s purpose. Both are waterways that create a large body of water. Reply @yoshikimatsui6384 3 years ago この分野は中国がいずれ大きな成果を上げることとなると思う。その成果は西洋の科学を超えてしまうかも？？？しれないと思う。 Reply @HPTrauschke 3 years ago Na super! Das hab ich jetzt gekauft und kann keine E-mails mehr empfangen. Ich schlage vor, dass ich nochmal 100€ drauf lege und komme bei Ihnen vorbei, damit Sie mir das bitte richtig installieren. Ich wollte eh schon längst mal mit Ihnen sprechen, denn ich bin auf der Suche nach jemanden, der das Wissenschaftsprogramm bei bluebox.earth leitet und da habe ich natürlich auch an Sie gedacht. Ich bitte daher um freundliche Kontaktaufnahme über die Website. Herzlichen Dank, HP Trauschke Reply @dirtdiggler9293 3 years ago It will never be completed! Nope not in a billion years! Scientists are not Nobel enough to Handel this accomplishment. Plus to what end Reply 5 replies @ElonizuMuskamoto 2 years ago You sound so German Reply @DallasMay 3 years ago The future of AI is really exciting, and scary, and the potential can't be understated. It will change people's lives in ways they can't even imagine today. Elon Musk is a guy everyone loves to hate, but he is the first AI Industrial titan. His companies are proof of the possibilities of AI engineering design. If you consider his engineering design teams, they are tiny compared to other competing companies. That's because the vast amount of engineering design in his companies is done by AI. His companies are making huge established engineering industrial companies completely obsolete, Boeing can't compete with SpaceX because the company of Boeing wasn't created around AI and the leadership doesn't understand how to use it. This is why Musk is the richest person in the world, just passing Jeff Bezos who got rich off of... understanding and implementing AI to make retail more and more efficient. If you think YOUR specific profession is untouchable by AI, you are wrong. Reply @tomashull9805 3 years ago (edited) Are you talking about the protein folding MECHANICS? Why didn't you say so!?! Nobody knows how self-assembly works either, like microtubules or bacterium flagellum... but quantum mechanics is most likely responsible... which can be tested by disrupting of quantum 2 Reply 3 replies @skipperg4436 3 years ago When will we learn to fold space?.. Sigh... I know the answer: "Never". 1 Reply @devilsolution9781 8 months ago Anyone else find her oddly beautifull. Reply @AbooRasta 3 years ago I am deeply disappointment that throughout this entire video not once did say 'Einstein' in your sexy German accent Reply @xavieryates9782 3 years ago Sabine, you are a beautiful woman and I want to marry you! Reply @thewaytruthandlife 3 years ago (edited) 2:12 Its weird that mutations lead to sicknesses ... according to the idea of evolution it should lead to improvements right ??? or are you saying that mutations dont support an idea as evolution ??? Well I can only agree to that one... mutations only lead to A) neutral effects OR B) sicknesses. Which makes sense if one knows what are the driving forces of mutations: in general mutations are pushed into existence by agressive destructive influences 1) internal radioactive decay of elements (C14 -> N14) 2) cosmic radiation 3) agressive chemicals 4) cross overs ( the only harmless form of mutations) 5) UV light 6) radioactive radiation from out side all of them are a) random b) have no purpose of enhancing living organisms c) has no intelligence behind it to push existing systems in to more advanced systems. c) are usually highly energetic Now random enegry inputs are a) destructive b) neutral at best. This concept is written down in another article of mine: www.academia.edu/38591420/The_4th_law_of_thermodynamics_docx Reply @EK-gr9gd 3 years ago Well, to be fair, it is quite easy to beat a Go player, in principle. It is not much different from the way Turing cracked the Enigma. The A.I. just needs to discard all impossible solutions. The tricky is thing is just "speed". Turing's "Bombs" needed several hours to work on a program. In a game of Go this may only take some seconds. Reply 7 replies @kingkhann9 3 years ago Oh for fucksake half the video doesn't speak of alpha fold Reply @ImranSahir1 3 years ago Is it me or she has a stupendous collection of clothes? And all of them exquisitely stylish. Reply @iankelly6632 3 months ago Yes #evolution did it Reply @user255 3 years ago My cells fold proteins everyday, not a big deal. Reply @PrivateSi 3 years ago (edited) I do prefer proper physics to quacky nonsense hypothetical 'theoretical' physics.. In this case the subject matter and processes are at least definitely proven to exist but the proof is still mostly theoretical, being A.I. based. This type of model, if implemented thoroughly, holds much more weight than say (Super) String theorists, Many Worlds literalists, Holographic Universe BS artistes, Computer Sim. worshippers, and of course Climate !Prediction! 'Science' fanatics (BS politics more like), as the factors are mostly known and there aren't so many compared to a Theory Of Everything or climate modelling.... Reply 5 replies @moses777exodus 3 years ago Q: How are proteins folded? A: DNA Reply @universalflamethrower6342 3 years ago Iuckily a-biogenesis and emergence explain this more easily Reply 6 replies @francissaffell6853 1 year ago It should have been solved by crypto currency. Reply @DumbledoreMcCracken 3 years ago I'm so sick of the artificial intelligence people 1 Reply @seanregehr4921 3 years ago Mankind is not even supposed to eat meat, so proteins play no significant role in the human body. The human body is electric. Meat plays no role in that function. Why eat dead food when the body needs living food to feed a living body? Do not be stupid. Man needs minerals for the body to function correctly not proteins, not vitamins, not supplements but minerals. The only proteins the body requires it makes plenty assuming you actually eat what is necessary to fuel the human body. Why are you promoting death by "eating"? Reply @kingkhann9 3 years ago Oh please stop stop stop.... I know you don't have any idea of biology or technology. Just make video of nordic vpn. Also how is Sabina spelt sabine? Reply @ianskinner1619 3 years ago (edited) fire the editor, these abrupt camera transitions made it impossible to watch comfortably.. 2 Reply @v.gedace1519 3 years ago (edited) Stopped watching after around 59 seconds. Why? Worst video editing in a while. Putting images left, than right to you. No/way to fast fade-in. Never seen this before and would not like to see it again. EDIT: The lecture doesn't seem to have a soul. The statements, which have certainly been thought about very carefully, are being thrown down. No emphasis, and pauses, really badly presented. Reminds me very clearly of some of my professors who were crammed with knowledge but couldn't impart it. Thank goodness, at least here I have the opportunity to decide for myself whether to look at it or not. I am out. Reply @osyris9281 3 years ago String theory is fake and i tis still a theory of ignorance, how can u speak about that and you think you have the big brain to explain about protein folding, think of the matrix , it is all about magnetism Reply @sanchopansa1950 3 years ago wenn man schon englisch sprechen zu müssen glaubt, dann sollte man es wenigstens können. grauenhafte aussprache mit vielen fehlern. Reply @ingemar_von_zweigbergk 3 years ago (edited) if I want to listen to lies, then I listen to a female older than 13 years old talk. Reply