basic science and translation you know the previous speaker described a phenomenal intervention but every day I um went to my hotel room and thought Oh my God I saw this one molecule or this one Target and it will change the world I mean so I'm I I I I I really get uh very enthusiastic about everything that's in the pipeline you know and um um I'll talk about methylation studies but because I'm the last speaker I feel I need to summarize some of my Impressions which are totally biased but we saw many talks on biomarkers of Aging for example Vadim gladish just presented this phenomenal talk on transcriptomic biomarkers and other biomarkers but there were also talks from Max unfred Brian Kennedy describing lipidomic and metabolomic clocks um of course we know they proteomics clocks and endless mulation clocks um as many people have pointed out um or I should say probably your head is spinning there's so many clocks what should we use you know and um the great advantage of um a proteomic clock or metabolomic clock and so on is interpretability if you really care about that inflammation I want to measure inflammation use an inflam um so the other thing is if you really want to predict future morbidity then use clinical biomarkers that detect early morbidity right so I have this topology biomarkers that measure early morbidity let's say blood pressure um are more predict predictive for future morbidity than an omix clock that Ma measures accumulated damage right so having said this the Aging field is very interested in biomarkers that measure ACC damage and I think it has a lot to do with our new understanding of the nature of um aging it's really accumulation of damage and disregulation um I want to start out with the main weaknesses of methylation clocks it is this issue of biologic interpretability I mean one can make biolog methylation Clocks interpretable by focusing on for example prc2 methylation Victorio talked about that so you can focus on methylation in certain chromatin regions um so that's the weakness but we heard phenomenal talk Talks by Raga seal and Ryan Smith that showed that these methylation clocks and methylation biomarkers are already being used in human clinical trials and I think the talk from Ryan Smith explains why methylation is such a strong predictor of future morbidity um and that's for me the key question why is it that these methylation clocks work so well why is it that you take somebody's blood samples and you can predict time to death and so on you know and what's the mechanism um and um then I think the field of course the biomarker field needs to integrate all these different layers of the M um and um biological hierarchy you know I think that will come um we need longitudinal studies in humans where we have methylation data protome data metabolome data and all sorts of omics um once you have longitudinal data with outcomes you can build wonderful uh models and um we heard talks from Rafa de Cabo and others that um describe these Enterprises of um collecting data but I want to come back to strengths of methylation despite all the limitations I think people like mation clocks because of an in intive appeal because it relates to to a molecule that carries genetic information we now know that this uh DNA molecule is also a clock that's just intuitively very appealing and also there's this intuition of the fact that the aging process is innate it's Universal everything ages all Maman species age all tissues age and so having Universal multispecies mulation clocks or pan tissue clocks appeal to that and then the fact another reason why they are so intuitive is that mation clocks are lifeour clocks you really can measure time from prenatal samples to centenarians and Vadim gladish just showed you how mation clocks indicate a very early Rejuvenation process that's a miracle of methylation clock said developmental processes are linked to aging much later in life so this is the intuitive appeal but there are technical advantages first of all I would argue methylation clocks are pretty much ready for human clinical trials many of you are already using these clocks clearly there will be better and better versions there were many Talks by different people who who describe new versions of epigenetic clocks and we will at some point have to have a an unbiased look and evaluation of these clocks maybe as part of this biomarker consortia that Vadim gladish um and Jesse and um um Marty mocky organize um there's another technical advantage of methylation clocks they actually work in vitro studies you know I mean clearly I love it that people use them in human clinical trials but the very same clocks can be used for studies for studying cells in a dish so Victorio Sebastiano showed how osm rejuvenates fibroblasts and other cells and then um my favorite reason for um methylation is the ease of developing pan melan clocks and there was this wonderful talk from Jesse Pani and the gladish cheef lab on heterochronic parabiosis we really learned so much from um the this Paradigm but what maybe escaped um many audience members is that Jesse used pan meleon clocks in other words whatever moved um this clock in the mouse model probably will also work in humans why it's a pan meleon clock so um talking about panon clocks um I call um these multispecies clocks third generation clocks and um a third gen one way to build a third generation clock is to um predict what I call relative age so what is relative relative to what relative to maximum life span so um if you are a 61y old human then your relative age is 61 divided by 122 which is the maximum lifespan of humans so a 61y old human has a relative age of 0.5 and similarly you can um um measure relative age in all species and the Miracle is how easy one can estimate relative age with meth with methylation and this brings me to another set of phenomenal talk uh Jackie and other speakers described um facial clocks extremely appealing and um from um Alex was one of the first people to um develop such a clock and I remember when I met Alex he he had an app on his phone and it nailed my age up to one or two years you know so um it's it's so such an appealing thing but my challenge all to the AI team is develop a facial clock for multiple species you know so um and start out with linking a human with a whale face and go to the dog and then a mouse and after you succeeded at that then put in the face of the axotal you know and and speaking of which there was a talk by Julia halua Joseph Zola maxun who did describe these dual species clocks dual species clocks that measure age relative age really for axotal and frogs so amphibians but also stunningly humans and amphibians you know so and that to me is yet another um Miracle of cytosine mulation it really speaks to my claim that there's this innate aging process that can be observed really in in so many different species certainly vertebrates so going back in time when I developed the pan tissue clock I wanted to find an intervention that actually resets it and in the only thing I could find in 2013 was the yamanaka factors IPS cells were perfectly young according to the pan tissue clock and and this finding is a gold standard IPS cells always are young according to all epigenetic clocks including panan clock so that's established but the field needs to sort out the issue about partial or interrupted reprogramming um because there things are much less clear um for example with there were great Talks by Victoria Sebastiano David Sinclair Luca koui Boris Georgia Vick many companies right now pursue that idea but we need to figure out um what dosage what kind of factors and in what tissues do we observe epigenetic Rejuvenation so partial programming has a weaker signal so now um I want to move on to a different concept maximum lifespan of species let's not talk about aging per se um so most people including myself really confuse species characteristics with individual characteristics so maximum lifespan is a species characteristic maximum lifespan of humans um 122 years maximum lifespan of a rat 3.8 years that's a species characteristic um however we um as individuals we actually only care about one uh characteristic time to death or time to morbidity that's an IND or or your age or your biologic age these are individual characteristics and I do want to tell you mathematically these two concepts are completely different but it turns out also biologically they're very different and the best example to illustrate it is um the oldest human whoever lived John cmore famous ly was a smoker so if you just copy her whatever she did you know um what I want to say is smoking doesn't affect the maximum lifespan of our species okay really doesn't of course it will affect your time to death you know and um however the longevity field has long been fasinated by molecular determinance of Maximum lifespan in species and we heard phenomenal Talks by Domino deer Vera gonova Yan VI many many people among the audience have studied maximum lifespan and um and me too you know I was fascinated by it and so I um assembled um a team the melan mation Consortium 200 investigators many people in the audience can find their name here um Vera gonova viim gladish Joe Takahashi many of you among the author list but together we assembl this large data set which is publicly available 348 species of different lifespans and we um so what were the key insights after seven years of pain and and here they are um biologic process processes that relate to time to death of an individual species what is that caloric restriction in a mouse um smoking in a human obesity all of these things that affect your a average health or lifespan they typically are completely different from the uh Pathways that determine maximum lifespan of the species so biologic age that everyone talks about is truly an individual property and so I had this question did I just waste seven years of my precious life to assemble a data set and it doesn't inform drug development at all and there was the best talk of the entire conference and I want to say the best talk ever in the longevity field by Alex zonov who coupled drama of science I could barely contain my excitement with with this amazing romantic twist at the end you know so I invite you um the Eternal longevity research Partners um Dominica vok and Alex anyways um I want to many people ask me about the miman mulation platform and mimon mulation studies and I want to tell you I started a non-for-profit um company which offers miman methylation clock testing um it's a fee for service Arrangement anybody can uh interact with them it has nothing to do with my current employer Altos lab so um I I personally like the melon methylation array for technical reasons but also all of our software is freely available for that anyway I want to talk a bit about some of the surprises of studying maximum lifespan with methylation one surprise to me was that when you do an upstream regulator analysis of these lifespan related cytosines you do see o four and socks too and some plur potency genes why is that interesting because it's again these yamanaka factors that biology strangely the biology of reprogramming seems to be linked to the biology of maximum lifespan interesting to me um more generally you could ask what is the methylome of a longlived species and here I show you um in red um the curve the the methylation landscape on the DNA of a long lift species it's hilly low methylation and transcriptional start sites high methylation in gene bodies high methylation in intergenic regions so hilly landscape longlived species by contrast a shortlived um animal has a flatter landscape and I think it has a lot to do with gestation time um the these animals uh are poorly built the methylation landscape is poorly built and poorly maintained so coming to another paper um predictor of Maximum lifespan we literally have a a multivariate regression model that uses a couple of hundred cytosines and that predicts um maximum lifespan on a log scale and so then we this is a biomarker and um personally I I think it's not a useful biomarker but Others May disagree but you can ask the questions which interventions affect the predicted value of the species maximum lifp span and I have the caution here this is about a species characteristic probably it doesn't matter for you and me in any event the first thing I looked at is of course interrupted epigenetic reprogramming right um and there's no effect you know so this interrupted cyclic reprogramming did not affect the epigenetic estimate of the maximum lifespan of the species it may still affect um the average life expectancy and I I was so happy that several speakers talked about uh clinical uh sorry Mouse studies where they do life span studies of uh transient reprogramming I briefly want to mention we also have predicted maximum lifespan for human cohort studies and our species characteris IC does not predict individual mortality risk in any of the cohorts we looked at another thing you can ask is what about cytosines that correlate with maximum lifespan do they correlate with chronologic age and the answer it should be right something that relates to maximum lifespan should change with chronologic age but exactly the opposite is the case um so there is no barely any overlap between um that so I cannot highlight that enough there's a distinct cpg pattern cpgs linked to chronologic age do not relate to cpgs linked to time to death in humans or in mice and this is deeply frustrating to me this finding um and so the hope is of course well what about if we form some sort of V diagram or what can we do and that brings me to a paper that is will come out in a week or so um which is about methylation Dynamics velocity gradient or whatever you call it I call it the rate of change in methylation so imagine you you have a longitudinal study you measure blood methylation from age0 to age 100 and you calculate a rate of change in methylation that rate of change will turn out to have a very significant correlation with lifespan um if you look at the right cpgs and so this plot is a little bit hard to read but each dot corresponds to a different chromatin regions promoters heterochromatin enhancers Gene bodies flank blah blah blah and the right hand side shows chromatin regions where the rate of change of methylation relates to one over lifespan Max and and it is this balent promoter regions and repressed by prc2 complex but the reason why I show this graph is there are many regions in on the DNA that don't show show that pattern so you need to be think about it how to define It Anyway here's here's the graph on the Y AIS you see log transformed rate of change of methylation every dot is a species and on the x-axis you see um log transform maximum lifespan beautiful relationship and you've seen this type of relationship a lot in your life how often have you seen it 10 times that's an old relationship people have shown the same for somatic mutations for tome for repair probably you can think of other things that should make you very nervous that you've seen it and um one of my punch lines is if you have a biomarker that correlates with age it will show this relationship I'll I'll come to that later anyways let's um talk about the Dynamics of methylation changes across the lifespan so by having the data from the melan Consortium we can look from Cradle to grave in pigs here and there's a distinctly nonlinear relationship of um gain of methylation at these regions and more maybe you appreciate it from here the rate of change in young animals is much steeper than in Old animals and that bears out so but interestingly the rate of change in young correlates with the rate of change in Old animals um I find that a high correlation speaking about correlations there's another interesting finding for for the quantitative types here I show how chronologic age correlates with methylation in different species each dot is a different species and on the x-axis we have lifespan and there's no relationship so um if I was a physicist I would say well maybe this correlation of age uh with methylation is a species in variant which explains another mystery of methylation why is it why can you build epigenetic clocks that are equally accurate in a shortlived species like a mouse as in a longli species like humans why is it always the same accuracy well this has a lot to do with that so um these fundamental equations that result from our methylation studies in mammals um fall into uh lead to several main insights the rate of change is inversely related to maximum lifespan in B veent regions but not in other chromatin States the correlation between methylation and age Bears no relationship to maximum lifespan and the rate of change in young animals is correlated to the rate of change in Old animals so um now I'm speaking to editors of journals and reviewers um because um coming back to my statement any biomarker that increases with chronologic age will res result in a strong inverse correlation between the rate of change and maximum lifespan and this is just the trivial consequence of the definition of the rate of change so one can derive I um the bottom shows you a mathematical correlation uh U um property where I didn't use any data this is math rate of change is one over lifespan times some factors so um one one really needs to address this bias in the definition you know um and I want to end by thanking my many many collaborators um Vadim gladish Vera Max y Kristoff Nas Jesse Pani Joe tagashi and all the people from the miman Consortium thank you so much thank you thank you thank you so much Steve that was that was a fantastic talk really amazing um we have time for a few questions um I will start with asking a question because I have the microphone in my hand okay um so have you looked deeper into the chromatin like like you could predict secondary DNA structures or something like that and whether or not there are changes in DNA methylation in those areas yeah I um well um I think many people look at that and um some talks talked about these R Loops I think that um I personally don't have anything to report on it but my instincts say that there is a connection yeah very exciting Steve um amazing talk um rate of change I like this second last the last slide that you had I kind of lost myself on it um it's a little profound um could you explain it again which one this slide can you explain specifically your lost equation a little more uh it's it's a little complex and I want to understand it and I think maybe a few others in the audience math and then people will guzzle down extra liquor to get it out of their system but okay the okay so assume you want to calculate a velocity a rate of change the upper equation shows you that you form a univariate regression model where your biomarker gets regressed on age the rate of change is the beta the beta is the left hand side of the equation so far so good then there is um that rate of change can be actually expressed in a beautiful formula I think which is that it's just the piercon correlation between methylation and AG divided by the standard deviation of relative age do you remember I talked about relative age is one over lifespan and that's it that's the equation it's a little homework exercise but I think I I hope that people do this exercise preferably today um because I'm really tired of reading papers on um this biomarker relates to maximum lifespan I'm just tired of it I think the reviewers need to ask tougher questions and say all right do some proper permutation test control for it you know because again this is without any data you know mhm yeah so SS uh this this turn oh yeah the this connection between Maxim lifespan and time to death for eptic data do you also expected for tpic and poic data um sorry whether we can build time to death estimator yeah you have got this disconnection between maim lifespan and time to death in terms of the am well vem should answer that you know my my uh I don't know I just think we need to be open to the idea that it could be difficult to translate interventions that enhance maximum lifespan um for us for health span purposes does that make sense you know um it could be that this is really um the goal of ardd 3000 in 100 years you know where they have species interventions yeah but yeah yeah I have a question regarding why why do you you think it's specifically the bent promotors that are correlated with maximum lifespan it's a biology question yeah that's an easy answer um um actually um so these bent regulatory regions that are typically Bound by the prc2 coms polycom repressive complex 2 they exhibit low methylation in pretty much all tissues but also all Mamon species and that part of the biology is one of the reasons why you can build pan tissue clocks pan mimon clocks you know so the polycom repressive complex represses but it does it without mation you know and um so yeah then you have low mulation and the only thing you can do with aging is to go up do you see so coming back to whatever goes up with aging will then have this um inverse relationship to Max lifespan thank you thank you Steve here H I have a question um Steve here yeah okay thank you thank you seems that there's something magic to dilation and to predict a biological age H is any connection the heterochromatin transposon silencing a chromatin organization because this is the function that dilation does at the end right maybe there's some changes that do do you can you speculate something have you seen something oh yeah I mean there are now several group groups have um for example developed uh epigenetic clocks that um measure um methylation in transposable elements line one I mean Vera has such a clock but couple of people and methylation I mean the the answer is there is of course the closest connection you know so and and people study it methylation in inhancer regions has effects you know so um everything is connected coming back to the actually the advantage of these blackbox pan tissue clocks because um if you allow a mathematical optimization method to capture it you probably capture a lot of these chromatin changes you know if you go in in a completely agnostic way you capture it all you know um but then you don't have interpretability you know so so the opposite would be built a clock that only uses line one mulation or only PC2 mulation or only enhancer mation you know so yeah hi Steve uh great talk thank you um I have a bit of a weird question so I heard Jennifer Duda discuss when she discovered crisper that she was excited but also kind of scared about the implications of her Discovery and you know we can now use these clocks to predict for example when people will die like with grimage is there any part of this that has scared you or that you're anxious about regarding these discoveries no I zero uh fears about biomarkers because think about it blood pressure is being used every day it's very predictive of mortality risk you know and people live with it yeah hi Steve great talk um so I have a question about uh of course partial reprogramming so you've shown that the current clocks don't really work so well with partial reprogramming and thus the slope does not change in the maximum lifespan uh clock right so yeah I wouldn't quite put it this way but maybe finish your thoughts yeah okay feel free to correct me okay so I'm just wondering if you think this might be due to the fact that maybe the partial reprogramming protocols that were used were not as optimized to detect such changes so yeah for sure for sure yeah so um let me rephrase things a bit so um so when you look at partial reprogramming long enough you know at some point you see the Rejuvenation how many days do you need you know that's the question but I agree with you um it's not optimized on many levels you know so probably that's one of the reason it really also depends on tissue type you know what I showed you though was a different question so I had an epigenetic estimator of Maximum lifespan not average lifespan big difference so so the species characteristic I I couldn't see an effect which um is unfortunately I would have loved to see one but I didn't you know yeah yes hi H hi Steve it's a very fascinating talk uh I'm Christine uh I'm L physician so I'm always dealing with human beings and the rid of Aging is quite a fascinating idea and I'm wondering for as long as we we we we want to calculate the rate how long should be the time interval for the test yeah that's a question um so there are so many commercial companies and wellness uh U clinics you know and they offer biomarker testing and the question is how often should you measure methylation uh I would not measure it more often then every six months possibly every once a year that's my opinion but I hope somebody will prove me wrong and then soon we have a smart watch and then every minute of the day you see mulation changes and um I want to finish with one thing um there was a study that claims um by um that there's a circadian rhythm talking about Joe tashi's work so apparently if you measure your epigenetic age at midnight you're slightly younger than at lunchtime so if you want to gain the system you know when to draw your blood that's why it's a great idea to go to Bar Seven yeah okay thank you so much thank you so much St amazing