ARDD 2023: The Mother of All Longevity Conferences Aging Research and Drug Discovery Meeting (ARDD)

ARDD 2023: The Mother of All Longevity Conferences This long-running conference was held in Copenhagen. Arkadi Mazin By Arkadi Mazin Sep 26, 2023 ARDD Crowd Once a year, Copenhagen becomes a Mecca for the longevity community. Hundreds of people flock to the picturesque Danish capital to attend the Aging Research and Drug Discovery Meeting (ARDD) for five full days of talks by geroscientists and biotech leaders, mingling and networking with like-minded longevity enthusiasts, and not-so-healthy late night-outs. Support Lifespan.io in the 2023 end of year fundraiser. ARDD is organized by Insilico founder and CEO Dr. Alex Zhavoronkov, Prof. Morten Scheibye-Knudsen of the University of Copenhagen, and their dedicated team of helpers who did a fantastic job putting this vast event together. Since 2019, ARDD has been held under the auspices of the University of Copenhagen, which provides its opulent Festival Hall as the main stage for the talks. ARDD is one of the longest-running longevity conferences, and this was its tenth incarnation. According both to dry numbers and the consensus of the crowd, this year’s conference was the biggest and best of them all, hosting some 600 in-person participants and around ten times more online viewers. It was also quite intense, with wall-to-wall talks, panel discussions, and poster sessions from 9 am to as late as 9 pm. In the final talk of the conference, Prof. Vadim Gladyshev of Harvard jokingly referred to those who had sat through all the 100+ talks as “centenarians” (just like with real centenarians, there were very few of them). Exciting and fascinating as all the talks were, we are only able to present a fraction of them here. Our apologies to those who were left out. ADVERTISEMENT Eterna is a clothing company with a focus on longevity. Proving the geroscience hypothesis In the inaugural talk, the veteran geroscientist James Kirkland, director of the Robert and Arlene Kogod Center on Aging, gave an overview of some aspects of the longevity field, including ongoing trials (mostly senolytic-related) and the search for good biomarkers of aging. According to Kirkland, more than 80 clinical gerotherapeutic studies are underway, interventional and observational, but biomarkers that reliably react to interventions are needed, which cannot be said about all current tentative biomarkers of aging. More broadly, Kirkland talked about recent research supporting the geroscience hypothesis, which predicts that aging processes start early in life and that targeting them should delay, prevent, and alleviate multiple diseases. We are seeing more and more of it – for now, mostly in pre-clinical models. As long as the underlying hypothesis is correct, we can hope to eventually find those elusive anti-aging treatments. Among the recent studies mentioned by Kirkland were a 2021 study that showed great results for the senolytic quercetin against COVID-19 infection (quercetin drastically lowered hospitalization, need of oxygen, and need of ICU) and two studies suggesting a “1-2 punch approach” for cancer, when cancer cells that had gone senescent following an anti-cancer treatment are subsequently eliminated by senolytics. So much is owed to so few MSCs Thomas A. Rando, Deputy Director of Stanford Center on Longevity, invited the audience into the amazing world of muscle stem cells (MSC) and muscle rejuvenation in the context of exercise. Muscle’s high regenerative potential, which makes it a good model of tissue repair and rejuvenation, is mediated by a small number of normally quiescent muscle stem cells that enter a proliferation frenzy when needed. Age-related decline in this proliferation capacity results in slower regeneration, more scarring, and stiffer muscle tissue. Rando reported on some of the recent work on rekindling muscle stem cells’ regeneration prowess. Exercise enhances aged muscle regeneration, and scientists are beginning to understand the molecular mechanisms behind this effect. For instance, exercise robustly upregulates the multi-role protein cyclin D1 in aged muscle stem cells. This negatively correlates with the expression of the cytokine TGFß, which plays a central role in inflammation. Both induction of cyclin D1 and nhibition of TGFß restore functionality of aged muscle stem cells. ADVERTISEMENT An advertisement banner for PartiQular supplements. Interestingly, exercise positively affects multiple types of stem cells, including neural stem cells. Its benefits are also seen in immune cells, such as muscle resident macrophages. This allows us to view muscle as an endocrine organ highly relevant to organismal aging. Rando also reported that plasma from exercised mice positively affects aspects of aging in age-matched non-exercised animals. On day 2, Rando delivered a second talk on epigenetic control of stem cell aging and rejuvenation. Aging results in dysregulation of the methylation of histones, the proteins that chromatin is packed around. Muscle stem cells exhibit dense heterochromatin. Since chromatin-related enzymes use metabolites as cofactors, the epigenome is basically modified by metabolism. According to Rando, S-adenosylmethionine (SAM) is required for the methylation process, as it serves as a methyl group donor. Polyamines, such as spermine and spermidine, consume SAM, and their levels increase with age. Inhibition of polyamine synthesis restores histone expression and loss of heterochromatin, and supplementation with SAM rescues the regenerative potential of muscle stem cells. The need for a strong grassroots movement In contrast to previous strictly scientific talks, Michael Ringel, managing director and senior partner at Boston Consulting Group, talked mostly about the economics of longevity and the importance of advocacy and education. Ringel began his talk with the largely undisputable maxim “Life and health are valuable”. Geroscience is a way to extend lifespan and healthspan for all, but amazingly, it’s “not yet on the public radar” – at least, not as the net positive thing it clearly is. To turn things around, Ringle said, the longevity field needs a strong grassroots movement. Today, he continued, conventional medicine is close to the limit of its abilities to fight age-related diseases, where spending more money results in only minuscule gains in life expectancy. Even conquering all major degenerative diseases, a hardly achievable goal, can only get us so far. Targeting the underlying processes of aging instead, which is what geroscience is trying to do, seems the only way forward. ADVERTISEMENT A link to a supplement website called NOVOS. ARDD Social Model In one impressive slide, Ringel outlined his view of the process of taking anti-aging therapies to the market. This, according to him, should start with influencing public opinion, which, in turn, drives legislation and resource allocation. Having advocated for more longevity education and lobbying for years, we at Lifespan.io cannot agree more. Ringel specifically mentioned two examples: Mary Lasker’s public awareness campaign, which led to the War on Cancer, and the more recent ALS Ice Bucket Challenge. The latter not only raised 220 million dollars in direct donations but also led to a four-fold increase in NIH funding, a 20% increase in research, and ultimately to three new FDA-approved drugs. Sadly, geroscience is still the stepchild of healthcare spending, with zero legislative initiatives, a meager 0.7 billion dollars in federal funding, and no FDA-approved drugs. The upside is that there is no way from here but up. Prevention is the key In his talk titled “From geroscience to gerotherapeutics”, Prof. Nir Barzilai of Albert Einstein College of Medicine echoed the previous speaker by pointing out that most of the increase in life expectancy humanity has achieved so far was due to prevention; hence, our goal is to prevent age-related diseases. To do so, future gerotherapeutics must target more than one hallmark of aging, as rapamycin does in animal models. Since the first day of the conference was partly devoted to longevity medicine, Barzilai, himself a medical doctor, touched on the relationship between geroscience and physicians. “From geroscience to gerotherapeutics,” he said, “we are on the right path. However, our biggest challenge is that there are many ways to target aging, and demand is outpacing supply. Many people could already benefit from maximizing their health, and having more clinics and physicians will pave a brighter path for all”. Barzilai puts a lot of hope in educating physicians on geroscience, but admits that the principle “do no harm”, taken too literally, “makes MDs conservative” when it comes to adopting the geroscience paradigm. Barzilai, the world’s foremost expert on metformin and the principal investigator in the upcoming much-anticipated TAME (Targeting Aging with Metformin) trial, gave an update on this drug. While things have sometimes been rough for metformin, with the Intervention Testing Program (ITP) failing to detect lifespan extension in mice, and one paper last year casting doubt on some previous assertions about it, Barzilai presented more optimistic data. According to him, people on metformin had half the mortality rate and hospitalization with COVID, which ties metformin to the hallmarks of aging. Discussing the appropriate time for a healthy person to start taking metformin, Barzilai mentioned the antagonistic pleiotropy principle, which suggests that not all drugs that are good for you when you’re old are also good for you when you’re young, and metformin might be one of those drugs. The bottom line is that metformin might not be advised for younger healthy people, although this demands further investigation. AI is the future One of the conference’s hosts, Dr. Alex Zhavoronkov of InSilico Medicine, delivered an update on AI use for drug discovery. Zhavoronkov described Pharma.ai, an end-to-end drug discovery and development platform developed by InSilico that consists of three modules: Panda.ai for omics analysis and target and biomarker identification; Chemistry42 for de novo small molecule generation and virtual screening, and InClinico for clinical trial outcomes prediction. According to Zhavoronkov, Panda.ai is already used by hundreds of companies and scientific teams. The era of generative AI is coming to geroscience, promising to speed up drug discovery even more while also saving a lot of money. This was made evident by some of the case studies Zhavoronkov presented. InSilico is also an active educator, helping create some of the best online geroscience and longevity medicine courses around. In his talk, Zhavoronkov announced a new course on disease modeling and target discovery, consisting of seven lectures and now available on the company’s website. Zhavoronkov was also visibly proud to present the new roboticized InSilico lab, looking straight out of a well-funded Hollywood sci-fi movie, thanks to elaborate ambient lighting and sliding doors. However, the big announcement came a few days after the conference: apparently, while in Copenhagen, Zhavoronkov was working on finalizing a major deal with the pharma company Exelixis. The latter has licensed ISM3091, InSilico’s leading candidate drug for treating BCMA-mutated cancers, for 80 million dollars in addition to undisclosed milestone payments. The drug is currently undergoing Phase I trials. Interestingly, among all the data-packed slides, Zhavoronkov also touched on the question of whether aging is a disease. In his view, it hardly matters in the context of pharmaceutical drug approvals, since “if the drug works in aging, it should work in a variety of diseases.” While there is truth in this, it is unclear to what extent this approach of hitting one disease at a time is limiting geroscience’s potential. Longevity medicine in the spotlight Longevity medicine, a budding field that aims at shifting the focus towards disease prevention in accordance with the geroscience principles, was on full display in Copenhagen. Prof. Tzipi Strauss reported on the world’s first longevity clinic affiliated with a public hospital, which opened in Israel earlier this year. Sheba Hospital has been on the list of ten best hospitals in the world for several years now, and it might be the perfect place for research into longevity medicine, as Strauss told us in her interview. Sheba is also a university research hospital affiliated with Tel Aviv University and running cutting-edge research. The executive medical board of the new center boasts big names, such as Nir Barzilai and Alex Zhavoronkov. In her talk, Strauss listed “the four A’s”, the principles that will guide the center’s work: Accessibility, Affordability, Academic (research), and AI. Israel’s healthcare system is among the best in the world, flexible and highly digitized, which should help the center to obtain various healthcare data and effectively run human trials. Tzipi reported on the upcoming SHARP trial (n=1500), the center’s first initiative, which will include several interventions that are rigorously tested using a battery of biomarkers. Another prominent expert in the field of longevity medicine, Prof. Evelyne Bischof, announced at the conference her moving to Israel to co-pilot the center. In her talk, Bischof reiterated the tenets of longevity medicine, including replacing the current “reactive medicine” that many people call “sickcare” with the prevention paradigm, which requires deploying methods novel to MDs such as multi-omic analyses. Longevity medicine starts with healthy longevity diagnostics and continues with developing highly personalized regimens. Bischof stated that it is imperative that high-quality longevity medicine eventually becomes available to all. Shealso gave a second, more science-oriented talk at the conference, centered on geroncology, where she argued for integrating longevity medicine into oncology decision making and guidelines. Dr. Andrea Maier of the National University of Singapore announced the opening of another hospital-affiliated longevity clinic, this time in Singapore. The new clinic, which opened its doors on August 31st, is affiliated with Alexandra Hospital and collaborates with the longevity center in Sheba (both centers are also collaborating with Mayo Clinic). Stressing the importance of biomarkers of aging for longevity medicine, Maier also gave an overview of this topic based on this fresh-off-the-press paper by Biomarkers of Aging Consortium, of which she is a member. ARDD Biomarkers Exercise and inflammaging Bente Klarlund Pedersen, of the Center for Physical Activity Research at Rigshospitalet, Denmark, gave a talk on the anti-inflammatory effects of exercise. Inflammaging, the ubiquitous sterile age-related inflammation, might contribute to as many as 50% of all deaths. It is known that exercise lowers inflammation, but the actual story behind it is much more complex and fascinating. Bente described the “IL-6 paradox”: while high resting levels of this inflammatory cytokine are associated with obesity and physical inactivity, acute exercise also induces a marked increase in IL-6. Apparently, IL-6 is released by muscle into the blood, which is followed by an increase in anti-inflammatory cytokines (sometimes up to 100-fold). This process is markedly different from sepsis, in which a similar spike in IL-6 is induced by an increase in another pro-inflammatory cytokine, TNF. While sepsis is basically an acute inflammatory response followed by an anti-inflammatory response, exercise induces a much healthier interplay of cytokines (probably by inhibiting TNF), which not just lowers inflammation but also increases glucose uptake, improving insulin resistance. Exercise also counters the increase in TNF induced by endotoxin. ARDD Exercise Inflammation Pedersen told the audience about a study in which healthy young males voluntarily and drastically decreased their daily physical activity for a period of two weeks, which led to impaired glucose uptake and insulin signaling, hyperlipidemia, loss of muscle mass and fitness, and an increase in visceral fat mass. The main takeaway is that physical inactivity induces chronic systemic inflammation (at least in part via macrophage infiltration triggered by visceral fat accumulation), while exercise lowers it. How do you want your gene spliced? Luigi Ferrucci from the National Institute on Aging gave a fascinating talk on a topic which clearly is not getting the attention it deserves yet: alternative splicing (AS), a molecular process that allows a single gene to produce multiple distinct messenger RNA molecules by including or excluding various exons or their parts in the final mRNA. AS evolved to increase the biodiversity of proteins produced by the same single gene, and, accordingly, it adds a significant layer of complexity to the genomic coding potential. AS is highly active during development and is relevant to determining tissue-specific gene expression. According to Ferrucci, alternative splicing is an essential component of biology that plays an important role in responses to such stressors as scarcity of energy (such as caloric restriction) or cellular senescence. Ferrucci’s group is also beginning to unravel the effect of exercise on alternative splicing, but, he admitted, “we are only scratching the surface”. It looks however that physical activity downregulates alternative splicing, which might be one of the mechanisms behind PA’s health benefits. Lipofuscin and macular degeneration In his talk aptly titled “Tales of the molecular garbageman”, Kelsey Moody of Ichor Life Sciences turned the audience’s attention towards another rarely mentioned player in age-related processes: lipofuscin. Lipofuscin is a yellow-brownish, lipid-rich pigment granule which accumulates in the cells of many tissues over time, especially in post-mitotic cells such as neurons and cardiac muscle cells. Due to the age-related nature of its accumulation, lipofuscin is sometimes referred to as an “aging pigment.” Lipofuscin granules are composed of a mix of lipids, metals (like iron), and proteins: a result of the incomplete breakdown of damaged cellular structures, mostly in the lysosomes. While the exact implications of lipofuscin accumulation are still a topic of research, it’s generally believed to be detrimental to cells. In neurons, for instance, it can affect cellular communication and might play a role in neurodegenerative diseases. In the retina, the accumulation of lipofuscin is associated with age-related macular degeneration (AMD), a leading cause of vision loss in older adults and a popular target of candidate geroprotective interventions. Inspired by research from SENS Foundation, Lysoclear, a company from Ichor’s portfolio, uses recombinant manganese peroxidase to counter lipofuscin accumulation in the context of AMD and has achieved some success both in vitro and in vivo. The company chose lipid nanoparticles (LNPs) as the delivery system, and the goal for IND-enabling studies has been set to spring 2024. Neural stem cells and lipids Anne Brunet from Stanford University talked about brain aging and rejuvenation. What makes this topic especially important is the brain’s irreplaceability. While we can hope to eventually be able to replace virtually every organ in the human body, this does not include the brain, for obvious reasons. This means that we have to find ways to rejuvenate it. The brain has some regenerative capacity due to several of its regions harboring stem cells. However, like in other organs, brain stem cells lose function due to age-related causes, such as transcriptomic changes and a decline in proteostasis. This leads to decreased migration of those cells to other brain regions, where they could have given rise to much-needed young glial cells and neurons. Brunet’s group focuses on lipidomic analysis of neural stem cells. According to Brunet, lipids, of which 50,000 to 150,000 varieties exist, are “vastly understudied”, and “little is known about their function, especially in the context of aging.” Most lipids that change with age in quiescent neural stem cells are complex membrane lipids. Interestingly, neural stem cell aging is accompanied by accumulation of PUFAs (polyunsaturated fatty acids, generally considered “good fat”). Changing lipid composition leads to less rigid and more permeable cell membranes and probably causes loss of stem cell function, but direct lipid supplementation has shown some promise against it. Fixing the matrix Sara Wickström from the Max Planck Institute for Molecular Biomedicine delivered a talk on another non-canonical subject: the role of the extracellular matrix (ECM) in stem cell aging. Stem cell niches are complex environments that actively affect stem cell health. As the ECM gets stiffer with age, it changes the niche’s mechanical properties. Those properties are critical for stem cell activation, as evidenced by in vitro experiments using hydrogels with variable stiffness. Scientists are not sure about how this happens, but one proposed mechanism is heterochromatin remodeling. Bottom line: old cells experience increased mechanical stress and remain in the quiescent state for much longer, which impairs tissue regeneration. Interestingly, old cells in young mice perform normally, thus lending additional support to this hypothesis. Another important talk on this topic was delivered by Collin Ewald from ETH Zurich, Switzerland. He talked about how a faulty ECM contributes to numerous age-related diseases. 333 ECM genes have been linked to the human “diseasesome”. In addition to this “matrisome”, Collin proposed the concept of the “matreotype” – a snapshot of ECM composition associated with or caused by a phenotype or a physiological state (such as health, disease, aging, or longevity). Restoring the healthy remodeling of collagen, the most abundant protein in our body, might be one of the most important goals for geroscience. However, ECM-related therapies are currently at the bottom of the geroscience’s agenda, with only 8 targets and 27 clinical interventions being studied, mostly in the context of cancer fibrosis. Collin also reported on a small-scale pilot clinical trial of DracoBelle, an extract from Moldavian dragonheads that boosts collagen production. Two-month supplementation led to marked increases in skin moisturization, elasticity, and density. David Sinclair takes the stage The third day of the conference saw a brief appearance of geroscience’s poster child, Harvard professor David Sinclair. Along with maintaining a high-visibility public profile, Sinclair remains a heavyweight researcher. Recently, his lab put out two important papers related to what he calls “the information theory of aging”. This theory sees loss of epigenetic information as the underlying feature of aging and claims that cells retain a “backup copy” of this information that can be used for rejuvenation. In the first study, the lab created a murine model of genomic instability via a mild increase in double-strand breaks (DSBs) in non-essential DNA loci. According to Sinclair, some elements of DSB repair, such as Sir2, also play an important role in maintaining chromatin stability; in essence, they are guardians of both the genome and the epigenome. How this “double duty” came to be is not clear, but it leads to increased epigenetic noise when this repair mechanism becomes too strained with age. In this study, Sinclair claimed, his team had been able to show that epigenomic dysregulation drives aging even when DSB repair is faithful and no harmful mutations occur, suggesting that loss of epigenetic information rather than mutations is indeed an underlying (or even the underlying) cause of aging. In the second, more recent, study, Sinclair and another Harvard geroscientist, Vadim Gladyshev, collaborated on partial reprogramming using small molecules. The study produced encouraging preliminary results, although much more work is needed before reprogramming with small molecules becomes as widely accepted as that with the Yamanaka factors. A while ago, Sinclair also made waves by restoring crushed optical nerves in mice using partial cellular reprogramming. In his talk, he reported on this year’s developments, which include a study that showed a similar improvement in vision in non-human primates. ARDD OSK Vision After answering the audience’s questions in a fully packed auditorium, posing for a few selfies, and signing autographs on his popular book “Lifespan: Why We Age and Why We Don’t Have To”, Sinclair was rushed to the speakers’ dinner, ending this rare in-person appearance. Aging trajectories, senescence, and cancer One of the conference’s organizers, Prof. Morten Scheibye-Knudsen, reported on his team’s work, which includes trying to understand various trajectories of aging using unconventional data such as medical records. These records, according to Scheibye-Knudsen, strongly correlate with age and reveal aging’s complexity and variability. For instance, aging is gender-specific, with males and females tending to have different trajectories (“males age faster but later, and females earlier but slower”, as Scheibye-Knudsen put it). Of course, aging trajectories also differ on an individual level. Organs and tissues also age in different ways: for example, the liver, according to Scheibye-Knudsen, ages in an almost linear fashion, while in the lung, aging progresses gradually (that is, steadily but not necessarily linearly). It appears that it is possible to predict age from a tissue’s (specifically, lung tissue) clinical features. Scheibye-Knudsen also talked about his team’s project that involves gleaning insights into possible anti-aging interventions by associating terms in scientific papers’ abstracts to aging. As an example, he brought up nintedanib, which showed a senolytic effect in fruit flies. Scheibye-Knudsen had more news on cellular senescence (a popular topic at the conference). In a paper currently published as a pre-print, his team shows a link between cellular senescence and risk of cancer. Senescence is known to have a complex relationship with cancer: it can be both a barrier to cancer and its promoter, depending on the context. By analyzing senescence-associated cellular morphological features in about 4000 non-malignant biopsies the team was able to predict the risk of developing breast cancer in the future. At the end of his talk, Scheibye-Knudsen announced the founding of the Nordic Aging Society – yet another organization that aims to help aging research and advocacy. Fasting without fasting Prof. Valter Longo of USC Davis School of Gerontology, a prominent geroscientist who rarely appears at conferences, talked about the fasting-mimicking diet (FMD), multi-system regeneration, and longevity. FMD was originally developed for cancer patients to recapitulate the beneficial effects of water fasting, which is hard to maintain. It differs from keto diets in several aspects, including low protein consumption and a less strict position on carbohydrates. Unlike keto, FMD is a periodic diet, not intended to be followed continuously. According to Longo, FMD leads to life extension and rejuvenation of the immune system in mice. It also appears to be beneficial for stem cell function by boosting proliferation and differentiation, including into insulin-producing ß-cells. In 2017, a study showed that FMD reverses hyperglycemia and prevents death in a mouse model of type 2 diabetes. FMD cycles can also reverse type 1 diabetes in a mouse model. What about human trials? Longo presented some promising unpublished data from a very recent one, but we are not at liberty to divulge it yet. Mitochondria: central to both apoptosis and senescence Among the several speakers who talked about cellular senescence was João Passos of Mayo Clinic. In his talk, Passos raised the painful topic of senescence heterogeneity, which hampers research in this otherwise promising area. While several senescence markers have been used in studies, not all of them are present in every senescent cell. For instance, both p16 and p21 kinase inhibitors are considered markers of senescence, but Passos showed that they have very different dynamics and can be expressed transiently. The senescence-associated secretory phenotype (SASP), can also differ considerably between senescent cell subtypes. ß-galactosidase, another popular senescence marker, can be present in non-senescent cells, such as activated macrophages. Inflammatory factors commonly associated with SASP can be produced by non-senescent cells too. Passos concluded that senescent cells should be identified via multi-marker approaches, including by spatial methods with single cell resolution, since senescent cells can be quite rare and far apart. “Single cell transcriptomics, proteomics, epigenomics will likely be the optimal way to detect senescence and its heterogeneity”, he said. The second part of the talk was devoted to why we need to understand mitochondria in order to develop senotherapies. Mitochondrial dysfunction, Passos said, is an often-unappreciated hallmark of senescence. Mitochondria are actually required for SASP production. A major mechanism behind this link is that mitochondrial DNA leaking into the cytoplasm through the membrane of dysfunctional mitochondria is recognized by the cytosolic-DNA sensing cGAS–STING pathway. This activates pro-inflammatory genes and SASP production, which can be prevented by clearing out diseased mitochondria. But why do dysfunctional mitochondria leak mtDNA? Apparently, it happens due to MOMP (mitochondrial outer membrane permeability) mediated by the proteins BAK and BAX. Originally, this is supposed to trigger apoptosis, meaning both apoptosis and senescence are regulated by similar mitochondria-related processes. Inhibition of the BAK/BAX complex suppresses mtDNA leakage and SASP production and improves healthspan in mice. Interestingly, stopping apoptosis is essentially the opposite of what senolytics are trying to do. According to Passos, it might be wise to go after the SASP while keeping damaged cells arrested instead of trying to nudge them towards apoptosis. The 106th talk Somewhat symbolically, in a closing talk or the conference named “Final Line, Final Talk: Defying Time, Defying the Clock”, Harvard Prof. Vadim Gladyshev went back to basics: namely, to the fundamental question of what aging is. He reminded the audience that among the 106 talks at the conference, very few were about aging per se. Is aging the continuous accumulation of damage or functional decline, or should we measure it via disease burden or mortality? Gladyshev’s view is that aging starts with damage, while almost the entire field of longevity biotechnology is currently focusing on late manifestations of aging. But how do we design experiments that measure aging? Gladyshev suggested that in order to target aging, scientists must identify multi-marker “signatures” of longevity (those associated with increased lifespan, including across species), of aging (those associated with processes of aging), and of rejuvenation (those associated with the rare events of transitioning from an ‘older’ to a ‘younger’ phenotype). Gladyshev is fascinated with the rejuvenation event (sometimes called “the embryonic reset”) that occurs during early development and allows an old organism to produce perfectly young offspring. Like many other geroscientists, Gladyshev thinks this event might hold the keys to understanding and eventually defeating aging. He announced a new paper from his lab on this topic, currently in press with Aging Cell. We will make sure to cover this paper as soon as it is published. Our nonprofit mission is to increase healthy human lifespan, for everyone! To do this, we need your support. Your charitable contribution tranforms into rejuvenation research, news, shows, and more. Will you help? ♥ Yes, I'll Donate CategoryNews  Tags:ARDD, David Sinclair, James Kirkland, Vadim Gladyshev Previous Post Next Post About the author Arkadi Mazin Arkadi Mazin Arkadi is a seasoned journalist and op-ed author with a passion for learning and exploration. His interests span from politics to science and philosophy. Having studied economics and international relations, he is particularly interested in the social aspects of longevity and life extension. He strongly believes that life extension is an achievable and noble goal that has yet to take its rightful place on the very top of our civilization’s agenda – a situation he is eager to change. Alex Zhavoronkov at ARDD2023: Aging Research and Generative AI as a Platform for Pharmaceutical... ARDD 553 views Oct 17, 2023 Alex Zhavoronkov, Insilico Medicine, Hong Kong presents at the 10th Aging Research and Drug Discovery conference: Aging Research and Generative AI as a Platform for Pharmaceutical Drug Discovery: multi-modal multi-omics and multi-species transformer-based age generators for target and drug discovery. Transcript Transcript Search in video 0:05 and we are ready for our next speaker I think you know him I can't I 0:12 can't remember where I've seen him before I'm extremely thrilled and happy 0:19 to uh present you Alex my good friend long-term collaborator whenever you're 0:26 ready please go ahead thank you uh and sorry I lost my voice completely so uh 0:31 it's been a little bit of a ride it's the fifth day of the event for some of us it's the seventh day so um very happy 0:39 to be presenting to you today I'm not going to be talking about our lab uh you probably have seen our 0:45 robotics facility that I've presented uh during the first day and I'm not going to talk about our software today uh I'm 0:52 going to talk about something that I'm extremely excited about uh and I spent 0:58 maybe 40% of my time doing multimodel generative AI so whatever you think is 1:05 going to impact our field the most I think that multimodel generative AI is 1:12 going to be one of those really top two or three things that are going to take 1:18 um our field forward the most so gener of AI is everywhere right now 1:24 everybody's talking about a Chad GPT you're probably using it right right now but it's not new so if you look back in 1:31 history and look at how it evolved so you can trace the origins of generative 1:37 AI all the way to you know Alan Turing people were talking about it for a long time but uh around 1:44 2014 Ian Goodell and yosu banjo published generative adversarial 1:49 networks that became so good that you could generate synthetic fake data deep 1:55 fakes that was pretty much indistinguishable from reality and that gave us um I'm not going to go deep into 2:01 those Technologies but that gave us the ability to um think even even even 2:07 bigger and uh um there were new technologies emerging and Transformers 2:14 emerged around 2017 with a publication from a team of uh Google scientists uh showing that now 2:21 you can use attention layers so the paper is called attention is all you need so if you want to read just one 2:28 read that one showing the the um importance of multi-head attention uh in 2:34 generative Ai and then you've got the explosion of gpts GPT stands for 2:40 generative pre-trained Transformer that's why this area is very important and where it created the most impact so 2:47 far was language natural language Pro processing but Gans variational outen 2:53 quaries were extremely important and are used today so people are perfecting them 2:59 improving them we are using them quite a bit um if you're looking at the evolution of open 3:05 AI GPT uh the first GPT was actually uh went online 2018 in June so you could 3:14 have anticipated what is going to happen so that a lot of people got excited there was a lot of hype but most people 3:20 just did not believe and um gpt3 went online you know middle 2020 so we're 3:28 talking about uh 3 years ago almost right um and GPT uh 3 and a half GPT 4 3:35 it's uh basically last year right so last 12 months which is pretty cool and 3:41 now it's changing everything uh GPT chat GPT is an extremely powerful Tool uh 3:47 there are many many many different large language models so there are different evolutionary trees of those models some 3:53 of those models are called Foundation models something that people uh build on top on uh they fine-tune them they 4:01 modify uh on top of them uh but and there are not that many of them usually 4:06 those foundational models are developed by very large corporations like Google Huawei um Microsoft uh open Ai and 4:15 startups that got a lot of money anthropic Etc takes a very long large 4:20 amount of money to usually build and train a foundational model uh and you're 4:26 going to have many many many of them uh going forward but first uh generative AI in drug 4:33 Discovery uh was actually presented 2016 2017 so my first paper on Gans for 4:40 generation of small molecules was published in 2016 we submitted June 2016 4:47 and then um December it was published uh and it was actually presented at ardd 4:53 which is super cool right so great idea to come to those conferences I'm going to since I lost my voice I'll let Alex 5:01 from the past to tell you a little bit of a about that good area for AI to 5:06 where is my voice let's do it again now let's see if 5:13 the AV team can put it through good area for AI to uh to disrupt and there are 5:18 two ways to two strategies for Aid driven Innovation and Pharma to ensure 5:23 that you get better molecules and much faster approvals 5:31 and another one is creating a new needle so our company developed a pretty comprehensive pipeline that utilizes 5:37 both uh of those strategies uh and uh we um uh start with a large numbers of 5:43 molecules large numbers of hypothesis and then take it through a process that gives us very high quality promising 5:50 leads that then later we validate and put into uh partnering and pharmaceutical drug development but 5:56 today I'm going to be talking about one uh technology in AI in deep learning called generative adversarial networks 6:02 which is essentially disrupting AI as well it's a new technology so yeah we presented on that 6:09 2017 and at that time very few people believe that it's going to work for a small molecule generation or generation 6:16 of synthetic data uh using generative adval networks or anything else like that and here by the way you've got one 6:23 very famous billionaire investor Jim melon um who invested in back down uh 6:31 which kind of gave us a boost and allowed us to forward and uh here is a 6:37 history of deep learning so many many Technologies were published uh before the con ultimate convergence uh in 2014 6:44 where ganss were published then there were multiple general purpose ganss here are the ganss for drug Discovery so uh 6:52 until 2018 most of my work was theoretical well 2017 201 17 uh in 6:59 November we started synthesizing uh and my first paper on um uh generated 7:06 generated small molecules and um selective J 3 inhibitor uh was published 7:11 in 2018 and then we had a big paper uh published in 2019 pretty famous made us pretty pretty 7:19 famous at that time um so in 21 days we demonstrated that we can synthesize uh 7:26 and test six molecules uh for an arbitary Target nominated by a partner 7:32 um that made us pretty famous so we did the work in 2018 took us a year to 7:38 publish almost got killed this paper and this is my my frequent collaborator 7:43 previously Fierce competitor Alana spur good from Harvard um and when this paper 7:49 got published we celebrated um and uh in this paper uh VY aptech challenged us to 7:55 develop a small to design a small molecule using generative AI for a very well-known drug Target called ddr1 kyes 8:03 we used a generative tensorial reinforcement learning system which uh is again plus uh the ability to uh very 8:11 rapidly reinforce uh synthesized six molecules four of them worked uh two of 8:17 them went into microsomal metabolic stability assays one went into mice all 8:22 is in 46 days that was pretty cool got a lot of attention uh and we turned uh 8:28 this technology into multiple tools that my team presented um uh I think a day 8:35 ago uh and now this tool is being used by hundreds of key opinion leaders this 8:41 tool is you know 10 out of the top 20 pharmaceutical companies are using him uh so it's a very very useful uh toolkit 8:48 and we also have a community around that and in clinico is a tool that is predominantly used by H funds and Banks 8:54 it predict predicts the outcomes of clinical trials uh and we use those tools to um develop a pipeline of 9:01 Therapeutics so of our own drugs because that's where most of the value comes from it's very rare for anybody in this 9:08 community to have uh your own Phase 2 asset from zero with Noel Target Noel 9:14 molecule um and this is actually going to be a pretty hot Target uh I hope it's 9:19 Dual Purpose aging and disease at the same time and now I want to kind of uh 9:25 bring you back to so some of you may have been at this event one wonderful conference by Vadim Glades and Steven 9:32 Horvath um last year mid last year uh and just before this conference deep 9:38 Minds uh Google's Deep Mind published a wonderful paper it's not as famous for 9:44 some strange reason but I was thrilled I was blown away it's called gate it's a 9:49 multimodal uh Transformer which took in multiple um data types and could perform 9:56 multiple tasks so it's not just one chatbot it could generate images it 10:02 could fold proteins it could chat to you uh it was not a huge model that's why it went largely unnoticed but I was blown 10:10 away that's me right there and uh at that conference I propose the hypothesis 10:16 in front of everybody so um vadam Glades is my witness uh one of the most 10:22 brilliant scientist who I deeply admire so my hypothesis was well why why don't 10:27 we actually trade uh a multimodel Transformer on multiple 10:32 data types and bring many aging clocks together with many different data types 10:37 so uh the Transformer could generalize and I call decided to call it even with an idea precious one GPT is 10:46 kind of like one Transformer to rule them all uh and here I'll need some 10:52 voice hi yes one clock to find the target 11:00 one clock to align them and in the depth of data bind them precious 1 11:07 GPT that's AI generated voice sorry we wanted to make it like very very uh Lord 11:13 of the ringy uh and um as promised uh we 11:18 published uh this year uh the work that we very very quickly uh turned into 11:24 action called precious 1 GPT multimodal Transformer based uh uh clocks 11:30 transcriptomics plus methylation plus a lot of metadata uh and it went largely 11:36 unnoticed because again you guys take time to pick it up right um and in this 11:42 model what we did we used uh different data types so in this case metalation and transcriptomics to train a 11:49 multimodal Transformer regressor to generate a multimodal clock that can 11:55 take both data types and and predict age and also identify the most important 12:02 genes that are um uh impacting the the accuracy ofed prediction and then uh so 12:10 here you can train on tens of thousands hundreds of thousands of samples and 12:15 transfer the weights onto a multimodal Transformer classifier trained also on methylation and gene expression um 12:24 annotated with disease so you've got a disease model that learns from AJ which 12:30 is a pretty neat idea and then you've seen how our Pand omics Works uh if 12:36 you've attended the presentation by Petrina and Frank and this system allows 12:41 you to pick the targets and prosecute those targets and very quickly annotate those targets again we teach a course on 12:47 target Sciences in silo.com can register and learn how to do disease modeling and 12:53 Target selection it's a it's an art and science for different indications different uh different different um 12:59 philosophies for picking the targets so if I were to give a really hard question to the previous speaker I would say okay 13:05 name top three targets that would be addressable by small molecules to postpone menopause that would be pretty 13:11 hardcore right and difficult to answer by the way but this system would allow you to um tackle those questions uh but 13:19 before I go into that so we of course look at the metrics for this clock and the metric sucked right so compareed to 13:26 B in class B standard clocks purpose built using different approaches uh you 13:34 know mean absolute error of uh 4.2 and metalation data is not going to be 13:39 impressing anybody right but here it's not about accuracy it's about the ability to combine several data types 13:46 and prosecute those Transformers for targets we also did a lot of other uh 13:51 different uh um experiments looked at how those networks perform on ipscs uh 13:58 on fetal development Etc but then we threw those networks into p omics uh and 14:05 as you know pics here's a list of genes uh that are usually uh that that's the 14:10 typical output of ponics at the end right that you're going to be massaging uh and originally we want to have low 14:17 novelty targets because if you are actually showing the target list like that to the head of the theraputic area 14:23 in big Pharma if they don't see what they already know at least to some extent they will just discard the tool 14:30 but then once you once they believe in it you can actually help them massage it to get to more novel targets uh to 14:37 satisfy their kind of ultimate uh choice of novelty confidence and Commercial 14:44 tractability so this system here you've got the list of genes um here every 14:50 column is a score based on omics on text Financial scores key opinion leader 14:56 scores here we actually see if you can trust the data coming from a specific scientist um and accessibility by small 15:03 molecules antibody safety Etc and we looked at the targets low novelty um 15:10 implicated in uh COPD and aging at the same time and we decided okay well there 15:15 are a couple targets uh that we can prioritize andics allows you to highlight them and then uh kind of 15:21 prosecute them work with a team around this target uh and we looked at okay ipf 15:27 aging and um IP F oh appin receptor pops up great uh and a few others right my my 15:34 target X is not here because we kind of masked it it's much more novel uh if you look at Parkinson's oh 15:41 my goodness appin ill when 23 receptor as well um and boom we look at heart 15:50 failure appin receptor scores stps and uh l23 um also scores STP right but there 15:59 are many others so again most of the time this is what you're going to see you're going to see major heterogenity 16:04 in targets so you are trying to um narrow down the patient population that is more likely to respond to your 16:10 treatment most of the time right so Target selection again it's an art and science but you want to find the targets 16:16 that could be used in multiple disease areas as a matter of fact our Target X 16:22 antifibrotic originally we found it for type two diabetes and also for aging and 16:27 then it looks like it works everywhere um an appin receptor many of you are familiar with that it's not a noal 16:33 Target uh it has been implicated in cardiovascular diseases uh it has been 16:39 implicated already in cognitive decline some big farmers have prosecuted with small molecules with 16:45 antibodies and now wait for it do you know one company that is taking this um 16:55 this target into the clinic well one company is called structure um they were called shouty we 17:03 were Neighbors on jlabs uh a long time ago it's company founded by Ray Stevens 17:08 which is which has glp1 receptor Agonist in phase two right so you know here in 17:15 this country uh you produce one of the most popular drugs that the world is addicted to and 17:22 everybody is injecting some glutamide glp1 U uh Target right into into their 17:29 bodies it's a biologic but if you can develop a small molecule drug for that it's super convenient right and you 17:36 don't need to have uh more very difficult production manufacturing shipping Etc so once this goes to the 17:42 market uh of course a lot of people will stop using biologics right but these guys they also realized that um you need 17:50 to have an appal and receptor inhibitor as well it's a good idea to Target this target usually as a small bio you don't 17:57 have a lot of money to go after many targets and they also prioritize appal and receptor my um so it's for uh 18:06 cardiopulmonary uh and see our our Target selection actually worked right so it looks like it's validated by this 18:13 guys have to bet their entire life on this right so it means that they have prioritized um uh the target uh for 18:20 cardiopulmonary and it's a good kind of indirect validation of pomix and 18:25 multimodal Transformer uh combo um and I actually think so again it's my kind of 18:31 wild bet is right now everybody is injecting seig glutamide and uh some of 18:37 those people who are injecting are even anorexic right so most likely uh you will have cardiovascular events down the 18:43 road and unless you figure out the way to build muscle right and protect cardiovascular system so uh these guys 18:52 have a very good strategy for for for for Target selection here but the front runner and here is a real kind of uh 19:00 really bomb to drop right um about appal and receptor is 19:06 bioage the wonderful chrisan Fortney who all of you know and if you don't know 19:11 you have to know her because her strategy is finding the targets using 19:16 longitudinal population data and then in licensing the compounds conducting phase 19:21 two trials augmented by biomarkers of aging and then taking it either to the 19:27 clinic or licens s it back to a big Pharma for a lot of money great idea right in the process you create a 19:33 commercially viable model where you can do aging research at scale and she is like one of the people I deeply admire 19:39 and her front runner where she's betting her life is appin receptor and again it 19:45 popped up using this multimodal Transformer classifier one built on 19:50 multimodal Transformer um regressor um trained on age uh and and receptor pops 19:59 up everywhere so there are a few other novel targets that we have right but if you are talking about non targets where 20:04 you can license a drug this is the one right and she licensed it if you go back 20:10 to the press releases I did it yesterday actually never prepare for the presentations unfortunately uh nowadays 20:15 when you speak to many uh too many conferences you kind of get to do that so I picked up this press release um 20:22 from um uh Business Wire uh and April 2014 2021 they actually licensed this 20:29 target from AMJ uh the the small molecule which is super cool um and by 20:34 the way deal like that usually takes you about 6 months is like the fastest you can do probably it takes like nine 20:40 months to a year so they've been in this business for a long time so um God bless 20:46 bio AG let's hope that they are successful and it looks like uh you know our model manage to pick up the targets 20:53 as well they have a different approach to Pi for picking targets but when most most many companies Converge on the same 21:00 idea and bet a lot of money on it usually there is something there right and I would trust this more than I would 21:06 trust the mouse model um so bottom line um what's next 21:13 so I don't have a lot of time to talk unfortunately but um so this is my road map this is what I'm extremely excited 21:19 about and extremely collaborative so now that we have published precious one GPT human uh now we are doing pre precious 3 21:27 GPT is done going to be submitted soon multiple mamal multiomics and it can 21:34 also generate synthetic data we have granted patents on this technology so we were one of the first to show that I can 21:41 use age as a generation condition for synthetic data generation so for example you take you know bread pits picture 21:48 extrapolate in time take one billion bread pits with different ages right and input conditions so you can do the same 21:55 thing um on transcript Atomic and metalation data and other omics so we 22:00 are using that using a multimodal Transformer um we have a new 22:07 architecture it's not the one that we presented in the paper and right now what we are working on as a Consortium I 22:13 brought in a couple uh collaborators for that um we're working on a precious 3 22:18 GPT that that can do monkey um uh human Mouse uh and dog hopefully uh many other 22:26 species will see uh planning to um uh to submit November it can also do data 22:33 generation Target Discovery and robotic Target validation so are doing it for uh 22:38 you've seen my robotics lab core data types Imaging methylation transcriptomics and imaging mation 22:44 transcriptomic response uh and precious for GPT uh will also be instructional so 22:50 you can talk to it and say well dear robotic lab now go and find some targets 22:55 that would work in humans and and monkeys and then we will chop some monkeys and put it in a robot and see if 23:02 it works um and now one other technology that I'm extremely excited about we call 23:07 it the reinforcement learning with expert human feedback what made uh chat GPT so powerful is actually not uh large 23:16 training we well training on a lot of data it's actually uh human reinforcement learning so they released 23:22 the system took tens of thousands of humans who evaluated the output of their 23:28 platform right to see if they like it or not that's why it's so good and it's so pleasant to talk to um in biology and 23:36 chemistry you cannot do that because you cannot get somebody for $2 an hour to evaluate whether they like the output or 23:42 not they will not understand it so you need to ensure that only expert humans 23:47 give you the feedback and what we did we um released the software to the community many of you are using our 23:54 platform and we don't take your data your data is protected whatever you are doing is protected all we like all we 24:01 ask you to do is like or dislike the output so we know whether the model 24:06 works or not and every time we get a case study from you where you are um 24:12 it's all gdpr compliant Etc you're doing it on your machine or on a private Cloud 24:18 uh but every time you give us feedback it goes back into the feedback repository and either Rewards or publish 24:25 punishes the models sometimes we kill those models completely if people don't like it if they don't work so now with a 24:31 large community you can actually train those models on Expert feedback at scale 24:36 so that g gives us a substantial competitive Advantage because nobody has this to my knowledge has U uh multiple 24:43 software tools in the field except for guys like you know open AI Microsoft Etc but you probably don't use them for 24:51 chemistry uh or biology uh and they to my knowledge they cannot be used for 24:56 that at this at this level so we have a pretty substantial community of experts that give us 25:02 feedback and uh let's collaborate uh this is my linkton uh as you know we are 25:07 very collaborative and very happy to be working with the wonderful groups like Mor sh nuts and lab so it's not only the 25:15 conference that they are organizing they're also doing Cutting Edge research in generative so thank you very much 25:23 happy to answer any questions 25:29 thank you Alex it's really inspiring to to hear about everything that you're developing we have time for maybe one or 25:37 two questions before the 25:44 break thank you for the amazing presentation you are talking in your 25:49 drug Discovery pipeline mostly about small molecules are you can you what about 25:56 applying it to other compound classes like small peptides or protein B we have we have a system called generative 26:03 biologics uh we released it you can get a pilot so now we can actually do small 26:08 peptides already so Alpha huses um with desired properties and one of the 26:13 properties that we optimize for is stability for example very difficult to achieve um and uh we allow other people 26:21 to use this system so good question um hopefully before the end of q1 next year 26:27 we release nanobodies and antibodies as well but we don't want to develop our own biologics so I think that if you are 26:35 good in small molecules you should be doing small molecules because that they are so much better right so you can eat 26:41 them uh you can ship them you can store them forever not forever but for a very long time uh they are much more 26:47 convenient than biologics but for the community we want to release the software yeah Alex um Rich you and I have known 26:55 each other for quite a few years I remember remember the hairstyle earlier on um yeah it was weird I didn't work 27:01 with Pharma back and I just have to I just have to congratulate you on all you've achieved not just through this conference but through your company too 27:08 so um if I think about normal course of speed course and speed in biofarma which 27:14 I've been involved in for a long time I see here in this conference some fascinating lead molecules whether 27:21 you're talking about autophagy promotion or senolytics or epidemic reprogramming 27:26 but I see two or three of them and two or three young companies so I think I can come back in 10 years and see that 27:34 they have got several molecules and the big farmer is beginning to buy one of them so then I come back in 15 or 20 27:41 years and they're on the market we have to change the normal course and speed 27:47 you have the capability to do that if we could link together what they have found in terms of targets and Lead molecules 27:54 and multiply it and move it forward fast then we have a field that would that 28:00 hopefully we can all see come to fruition I'd just be interested in your thoughts about uh the statistics and the 28:06 time course sure well what I figured out now we work with many many Pharma 28:12 companies and there are many farmers here so what I realized is that uh the 28:18 pharmaceutical industry is deeply broken so recent estimates by Alexander 28:23 Schumacher and Oliver Gusman demonstrate that the cost of drug Discovery and development is $6.1 billion per drug if 28:32 you do it within big Pharma if you do it within if you account for small for for small biotechs it is um uh about maybe 2 28:40 billion if you kind of uh average it out but it's still huge right so it's huge 28:47 uh and usually it takes about 10 to 12 years uh also what I figured is that you 28:52 can go to preclinical candidates so one notch before clinical so then you do the enabling studies and go clinical so to 29:00 get to preclinical candidate I can do it in 9 months and I've demonstrated that once already so my longest one is 18 29:08 months my uh shortest one is 9 months and with generative the level of quality that you get for the molecules is 29:15 actually higher than what they got used to the big Pharma um companies or even like Japanese biotech when they look at 29:21 the molecule like how did you do that um and uh um 29:27 getting to preclinical candidate is easy and here you will have a lot of candidate molecules getting through 29:34 phase two is extremely hard and you cannot accelerate that so to change this 29:39 we actually need to figure out if some country or some regulator would be 29:45 willing to work with for example us and there are probably I don't know one other company like ours in the world uh 29:51 to try to set World Records kind of it's all about figuring out how can you set 29:58 the speed limit by showing that you can set a record right and um there we will 30:05 actually ask for more oversight not less but I would want to work with a specific 30:10 cohort of patients demonstrate that we can very quickly get to proof of concept 30:17 maybe less number of patients uh but the FDA needs to be convinced or whatever 30:22 regulator we would work with um and then also at the same time measure a lot of 30:28 Aging biomarkers getting this Aging biomarkers in a regular Phase 2 clinical 30:33 trials it's actually not easy and what your chief medical officer will be telling you is that don't do that don't 30:39 collect this data because if you see something weird you're going to be you're going to you're going to die 30:44 right because uh phase two failure in our stage for example is death right or 30:50 at least like 30% 40% uh down and you lose 10 years of your life and a lot of 30:55 investors money so you want to have clean studies right that achieve their goals and then you expand so I think 31:03 that to change the world we need to have The Regulators working with us hand in hand and you need to have very dedicated 31:09 regulators and right now those dedicated Regulators are do not exist in Democratic countries right now right 31:15 because they are like everybody will have a say and it it just doesn't work 31:20 so we just need to do what we do uh gradually right there is a really really 31:25 nice um say I think in Chinese um now I I won't be able to remember I 31:33 don't speak Chinese uh but basically step by step we'll get to the goal um 31:41 um ah I 31:47 I great um it has like multiple swear words in that in other languages so it's 31:53 easy to remember uh but yeah it's we should definitely collaborate work together and step by step reach the goal 32:00 thank you so much Alex that was a great ending [Applause] 32:09 point Follow along using the transcript. Show transcript ARDD 5.47K subscribers Videos About 3 Comments rongmaw lin Add a comment... @wichetleelamanit6195 @wichetleelamanit6195 1 month ago Thank you very much for sharing the knowledge. Reply @Superlongevityinstitute @Superlongevityinstitute 1 month ago Wonderful 😊 Reply @joeschmoe5583 @joeschmoe5583 1 month ago OMG we're finally getting ARDD videos - for the love of God - when is DAVID SINCLAIR going to be uploaded? Thank you!!! 1 Reply