49:16
08/14/2024
Episode 7

The Power of Single-Cell Epigenetic Profiling

In this week’s episode of the Everything Epigenetics podcast, Dr. Hannah Went sits down with Alex Trapp, a computational biologist deeply involved in aging research. The two explore the cutting-edge development of single-cell epigenetic clocks and their potential applications in the field of longevity science. Alex shares his journey from experimental biology to computational approaches and discusses the significant implications of understanding aging at the single-cell level. He also provides insights into his work at Retro Biosciences, a company focused on extending healthy human lifespan through innovative therapies.

In this podcast you’ll learn about:

– The development and significance of single-cell epigenetic clocks
– Differences between bulk tissue and single-cell epigenetic sequencing methods
– Applications of single-cell aging clocks in understanding and potentially reversing aging
– Challenges and future directions in single-cell epigenetic research
– The exciting work being done at Retro Biosciences, including advancements in T-cell rejuvenation and plasma therapeutics

About this Guest

Alex Trapp

Alex leads the computational biology group at Retro and previously developed a single-cell epigenetic age profiling framework at Harvard Medical School. He studied computational biology at the University of Rochester, focusing on the naked mole-rat’s immune system.

Twitter/X

Google Scholar

Alex’s Study

welcome to the everything epigenetics podcast where we discuss DNA regulation and the insights it can tell you about

your health welcome to the everything epigenetics podcast Alex super stoked to

chat with you today yeah I’m happy to be here thanks for thanks for hosting of course I I

want to you know I just want to know more about you you’re you’re so interesting you have a great background

and i’ love to have you explain to listeners you know how did you get to

where you are today can you just tell me about your your journey and what led you here yeah sure so I was actually born in

France uh the south of France uh moved around quite a bit when I was young and then uh ultimately ended up in the US an

elementary school uh I’ve been there pretty much since then went to college

at University of Rochester in Western New York uh really cold place but but great um really enjoyed my time there

that’s kind of where I got into the Aging space to begin with uh I worked with Vera gorbin NOA who’s a comparative

biologist by training uh he was really interested in the naked morat which is the longest living rodent and I also became very interested in in this animal

I lives more than 30 years uh which is incredible for an animal of that size

and started out doing mostly experimental work a lot of cell culture I was really interested in the himat totic system so um basically the the

entire immune system and the how the hemat totic Sy become every other type of blood cell in the body um and then

yeah I think I slowly transitioned more computationally uh I studied abroad at University College London first semester

took my first computational class there and I realized there was just a huge amount of power to to computational

approaches in modern day biology so came back switched over uh that’s where I first got exposed to single cell RNA

sequencing and a couple other single cell methods which is I think what I work on now as well um and then yeah

after that graduated in May 2020 and then moved to theing Glad Chef’s group at Harvard which was uh a lot of fun I

was there for a year and a half um yeah viim viim is a really incredible mentor

and everyone in the lab is is great I think he just talked to Jesse Pagan recently as well from the lab so it’s a

yeah really great place to study aging and really kind of the fundamental questions that are that are still unanswered when does it begin why does

aging happen how does it happen and really taking kind of a computational look at at some of these things um and I

specifically onited clocks uh published this paper that we’re talking about today and as well as as a couple couple

others and now yeah then I moved to to Retro in January of 2022 I really wanted

to I think Academia is great but I really wanted to translate some of these findings um into something that can be

used in the clinic I think that’s it’s a very powerful driving force for me uh to really improve the lives of people and

and yeah it’s been it’s been an awesome Journey uh building the computational team here and and we’re really growing the company quite rapidly as well so

it’s been it’s been a it’s been a cool Journey for sure that’s great and we’ll talk about retro at the end for everyone

wondering what’s retro I want to know more because I think you’re right you know some some people love the Academia

based side and they’re very dedicated to that but um it’s we don’t get a lot of

Alex Trapp’s journey in the aging research field

um what am I trying to say more of uh yeah taking this to to clinical use and in kind of clinical practice so super

excited to hear about your your experience there and you’re right we’re going to be talking about your great

paper the profiling epigenetic age in in single cells but we’ll we’ll chat about you know everything here and there um

and starting with the big picture you know before your paper was published in nature um in March of of 2021 all

existing epigenetic clocks relied on measurements derived from bulk samples so meaning and you can clarify this

further but basically samples that contain many cells you know a saliva sample a blood sample a buckle swab um

but you’re taking a different approach you’re going to be looking at single cell epigenetics so what the heck is

that um what caused you to do that why is this important give us all of the the

goodies yeah so so as you mentioned um I think basically most of your listeners are familiar with this concept of

epigenetic clocks of tracking aging biological aging through through the methylome can also be done with any

other modality right it’s not exclusively the methylone you could transcriptomic clocks you could build proteomic clocks these are some of the

things that we’re also working on at retro but um yeah to get to your question I

think there’s something cool about getting like an age metric from a whole tissue like let’s say you just give a

vial of blood and you get your biological age for your whole body or maybe just for your blood but there’s something also very interesting about

how individual cells age and I think that the last uh specifically the last

decade I’ve seen this uh this this huge increase in interest also in Technologies uh to to measure single

cell uh and you can do this across the epome transcriptome and they are now multimodal assays as well that allow you

to measure multiple modalities in the same cell which I think is is really incredible and and the beauty of single

cell is really that you see the heterogen that occurs in any tissue so this is particularly useful in tissues

that have a lot of heterogen so you think think about the blood there are there are countless numbers of cell

types all doing a bunch of different things and if you could understand how each of these individual cells age uh I

think it’s just a really interesting kind research question and also has a lot of of applications um so that’s what

kind of got me interested in it um I think there’s obviously a lot of value still to these bulk measurements and

bulk clocks I think they’re probably going to be more accurate for for a while but especially in uh especially in

a therapeutic setting when you’re trying to figure out kind of perturbations or or things that you can do to cells to rejuvenate them you need to be able to

measure the age of those individual cells and so that’s what really got me started into the into that

Journey yeah and was that how how did that idea start in the lab um like was

that was that you was that I know V Dr gladish just there are so many people I’ve connected with already um after

creating everything epigenetics it’s like you know you start making all these connections and it’s just fascinating so

like how do you even come up with this idea um yeah this is definitely something that the vium was really

interested in uh when I joined the lab um and I also got very interested in it uh after talking to him before before

even joining uh we had a couple Zoom calls and and he kind of threw around this idea and I was like oh that’s

that’s very interesting like why why has no one done this before turns out that there were a lot of complications right

so there there was really minimal amounts of data to begin with of public data to use and it just required a

totally new algorithm to work in the first place but I think it was a really interesting project because it really

Bridges kind of this world of epigenetics and aging which has just um exploded over the last couple years with

also some of these new technological advances that again enable you to understand heterogeneity and and really

measure things at the the fully functional unit of life which is a cell right so yeah but I I definitely need to

give adim a lot of credit for it I think he’s a he’s a Visionary and he’s been uh yeah very very impactful in my my

development as a scientist for sure sure very exciting to yeah hear hear about the work that you know has has come out

of his lab and and continues to come out of his his lab as well so when we’re we’re talking about the epigenetic age

clocks like you said most most people they’re they’re getting them from these bulk tissue samples um what are the differences then between the output of

Overview of epigenetic clocks and single-cell approaches

the bulk epigenetic sequencing methods and then the Single Cell sequencing methods that you did and and feel free

to elaborate spend a lot of time on this because um again I think this is very foreign to listeners and and really new

so I want them to be able to to understand this yeah for sure so um

most most methylation uh profiling tools um in the past have been again in bulk

and they just use these arrays so it’s a microarray you hybridize some some DNA

onto that array and whether it’s methylated or not you’ll get a different reading and so you can measure methylation based on that but basically

all those individual reads that you’re actually hybridizing to the array come from a bunch of different cells um and

so what happens when you’re doing the the downstream data processing is that you get a fractional value of methylation for every site so every cpg

you’ll get a value between 0 and one and that’ll tell you basically how methylated that cpg is across a bunch of

different cells in your sample and the nice thing about that too is that you start with a lot of DNA if you have thousands or millions of cells in your

sample you’re going to have a lot of DNA so you’re going to be able to cover consistently a lot of cpg sites and

that’s what most traditional clocks relied on they they relied on you take this array you have values for each of

these cpg sites and you’re able to train these kind of classical shallow machine learning models like elastic net um that

that require kind of all the these sites to be covered consistently um and that’s great I think but in in methylation at

the moment at least it’s really difficult to consistently cover all these sites so you deal with a huge

problem with sparcity and this is common across all single cell modalities um there’s imp being it’s improving all the

time but there’s still a huge problem of disparity so not only do you get different sides covered in in every

different cell but also the the the nature itself of of the output is different so again in bulk it was

fractional so you can have like 30% methylation 50% 60% but with single

cells it’s theoretically more of a binary signal right because that cell should just be on one strand it should

be either methylated or unmethylated and you have exceptions obviously you can have heem methylation where you have one strand that is methylated and one that

isn’t but in most cases it’s going to be a binary signal so kind of dealing with the sparsity and the binary signal um

are are the biggest differences uh but there’s improvements I mean this was in my paper I mostly analyzed single cell

whole genome bisulfide sequencing but that’s what produces the random reachs across the genome but if you do it in a

more targeted fashion and there’s been a couple really cool preprints recently that kind kind of explore this idea I

think you could theoretically have a fairly similar um fairly similar Matrix

that that what you would have in bulk so it should enable even more methods to be a to be developed uh in my space sure

sure that’s it’s just amazing yeah I I like all of the um Graphics you use in your paper too it makes it really easy

to understand so for everyone who’s thinking what the heck is going on I’ll put those graphics and maybe like some

explanations and stuff and of course site your paper I hope everyone um listening who’s who’s interested in

wants to learn more actually gives Alex’s paper a read because again it’s it’s fabulous this is just so Cutting Edge and really new stuff um where where

are you getting these uh samples from like the single cells can can you give

me a little bit of background on that and how that works yeah of course so we were actually really lucky and that

there was like not a huge amount but enough public data to to have a proof of concept kind of study uh which is

ultimately what we did so yeah we just used purely uh publicly available data there was one data set from Yan’s group

uh that looked at hytes so liver cells there was another great one in in muscle stem cells um from W fr’s group uh W

who’s now alos slads um and he also has this really cool studies on renesis

which is I think the Crux of the paper in my opinion um and also some of these embryonic stem cells as well so but yeah

we were lucky to be able to mine kind of the public database but I hope that more and more people and and this is something that I think a lot of people

are interested in is just generating more of these single cell profiles higher quality sequencing uh just to

understand more and more about these cells yeah and um we’re chatting about

just that the the remarkable progress that there has been in these these single cell omic kind of profiling

methodologies I I guess you can call it so in in your paper you mentioned there’s still some some issues some things that need to be figured out there

um Can can you describe some of those issues and then how you you overcome some of the the limiting factors or or

what you ran into because at the beginning you said you know yeah this is going to be really hard to do you you ran into to some key points yeah yeah

for sure um so I think this kind of ties back to to one of your previous questions about like what’s different

between bulk and and single cell and and here again it’s mostly this this binary nature of the data so having to deal

with zeros and ones as opposed to a fractional value and also just the sparsity right so in one cell you might

have one set of cpgs covered and in another cell you’ll have a totally different set and this is problematic

for kind of these conventional uh machine learning approaches that people had employed to build up genetic clocks in the first place so the way I tackled

that was uh on one hand for the sparity issue is essentially like like

pre-computing and training models on on bulk methylation to to kind of map out

how cpgs change in bulk across the AG strange and this is all in mouse in mice

uh this paper although I hope ultimately this is applied also in humans that’ be really cool um and so you can kind of

like map out the trajectories of these cpgs and since you have kind of this pre-computed set you can intersect that

set with any new single cell that you get and only only look at the cpgs that are covered jointly within both of those

sets so you’re not um trying to kind of impute information of any kind I

explored some of these approaches but they also aren’t super mature at the time uh maybe maybe in the future they

will be but so now B you basically only use the data that’s available and you map the Single Cell to this like

reference data set of aging and I built these in a tissue specific manner so I built one for the liver could also do

multi-tissue and it worked as well um although a little bit less well because it’s trying to learn um kind of these

tissue independent aging Dynamics which I think are harder than tissue dependent ones and so that that was for the

sparity and then for the I think this also kind of solve the the binary problem cuz basically if if you have

more methylation in an older sample you expect to get more individual single cells that are methylated in that older

sample if you were to sample it again right and then conversely if it goes down and so you can kind of track how

this how this goes and then estimate kind of like with with a probability how close how how old um it it is basically

could compared to yeah there’s there’s a lot more details in the paper but essentially

yeah yeah yeah you just like yeah um yeah you you map the the

binary methylation level with this this probability that you compute um and on

for a single cpg that doesn’t work super well but when you aggregate over many age related cpgs you’re able to build

these really nice kind of gaussian normal distributions and then the top of that distribution the maximum likelihood

is is ultimately the age that we kind of predict is the age of the cell uh but obviously there’s a lot of like

technical variations things like that so it’s it’s not a perfect method by any

means I mean if you look at some of the graphs right the Box blocks are pretty big it definitely shows differences

between young and old and I think some of this embryogenesis stuff like gu is very significant uh but obviously a lot

of improvements remain and there was recently a paper um another preprint by wolf Reich where they kind of try to

improve a little bit on my algor algorithm and then add some kind of future selection steps as well and so I think it’s it’s really cool to see

people contributing kind of improving on on methods because I think yeah in one way this is kind of the future if you’re

able to understand how individual cells age and you’re able to make really good metrics for that you can totally start

to evaluate a whole range of Therapeutics that you otherwise wouldn’t have been able to to look

at it opens up a like another like it’s still epigenetics but

it opens up an entire different field of of epigenetics like that hasn’t even been touched so I mean again that’s why

was so excited to to speak with you so um no I appreciate you you really diving into the methods there I know it’s it

gets a little bit more technical so again people can read through that if if they want to learn a little bit more and you said this was done on on mice right

remind me yeah so this is the the whole paper was done on mice but theoretically you could do the same with humans so if

you’re able to make these kind of reference data sets in humans and then also get single saw data in humans which

there is um there’s much less single cell methylomics data than there is transcriptomics I think single cell or

NAC because by far are kind of the Behemoth in the space but now there’s also a huge amount of interest in single

cell tax seek method developed by William Greenleaf and Jason bun Ruster at Stanford who now has his own Lab at

Harvard as well um and S said these like multimodal assays too I think are like super exciting right if you can measure

multiple kind of layers of biology in a single cell it allows you to start looking at like temporal Dynamics and

how like changes in the chromatin actually impact changes in transcription and I think there’s also huge interest

in proteomic right so the three layers the DNA the RNA and the proteins if you can measure all three at the same time I

think you get a very comprehensive view of the cell that you were never able to get before and I think ultimately this

will this will be really applicable to aging and any kind of Rejuvenation research is understanding those multiple

layers and building models and and testing your Therapeutics or your compounds or your interventions on these

and being able to evaluate them for the first time yeah fascinating just yeah

extremely interested in again everything that you’re doing um so so yes let’s

keep let’s keep diving right into it um this has been been great so far Alex so you you basically developed this um is

Development and challenges of the single-cell aging clock

it just SC AG single cell age right yep SC is the name yeah SC so this this

novel clock um method capable of profiling those epigenetic age in single cells so I think you talked a little bit

about how you did this and and how you designed it is there is there anything else you want to add in terms of of maybe the creation behind

it yeah I mean it was a it was a really fuus kind of story um at the time in the

lab I was doing experimental work um I was working with a postdoc or instructor

Marco who’s now a professor in Spain and we’re really interested in seleno proteins which is also a big interest of

vs I think he’s kind of pioneered that space in many ways and to use some of

the the recoding mechanisms that are inherent thear protein biology to try to address uh uh like stop stop Cod like

stop Cod on mutations and then trying to read through basically uh those mutations and these are involved in a

lot of diseases in humans uh and so it’d be really awesome if you’re able to kind kind of cure them through this seleno

protein um yeah like through harnessing seleno protein biology and using kind of these

uh other approaches uh to to solve that but at the same time I was getting more

computational a little bit I want and I think I got to the point where I had a nice proof of concept for this experimental um experimental project

that I was doing and and then I was chatting with adim and he he was like yeah I think this single cell stuff

would be really important to do so then I kind of switched over fully to this uh the single cell and I toiled for a

couple months uh with very minimal progress I was trying all the conventional approaches trying get some

of these imputation methods I was trying a lot of things and nothing was working I think this really speaks to like High

resilient one needs to be in science because for for a long time things don’t work and then um what happened is that I

I got an email in in February of 2021 from one of my closest colleagues at the

time chabba who’s a was a post do talk from Hungary at the time is not professor in Hungary and he sent me a

paper from a from a group in Germany by Wolf Gang Vagner group and I read that paper and and the story behind that

paper is also really interesting because it’s a paper that de had sent to another colleague of mine Patrick Griffin who works with David Sinclair and then

Patrick had sent that paper to chabba and then I had received the paper from cha um this is a late at night I

remember on in February it was really cold in Boston and I read the paper and I thought it was uh really interesting

and they kind of had some interesting kind of algorithmic ideas in there trying to do some of these predictions

based on like the lower number of sequence reads which is ultimately what single cell is also kind of like

compared to bulk and I read it and I was inspired and that that very night I could at the very

first implementation of so it improved over time but the very first one was was there that night and I remember emailing

viim the nights then I was like viim I think I think I have something and so yeah and then like I I worked really

hard and and got a preprint out in in about a month uh which was a marathon for sure uh then yeah the review process

lasted a couple months and ultimately it was it was published in December of 2021 now so yeah it was story yeah very very

thankful for everyone that was involved um yeah the perseverance like I my I I’m

learning a lot more about the computational um you know biology and the bioinformatics and and statistics

and everything that goes in it that’s that’s not my background um so I’m learning as as time goes on and

definitely not even close to being an expert in it um trying to train myself and and learn more but just the type of

volumes it speaks about who you are and and like and people who who take that up and and dedicate you know their life

work to it because you’re right like if you run into this problem and and your coding and and I mean you you can’t

figure stuff out it’s it’s just like you you know it makes sense to come together as a community and and have all of these

codes public and and um you know kind of the moment when it clicked for you that just that feeling has to be so good when

it like actually happened and you sent that email I’m sure that was like no other feeling yeah it’s definitely a beautiful

thing CU for a while I was getting not the greatest results and uh but that

happens in science right like it’s usually usually work a while and then sometimes there’s a little bit of a

breakthrough and then things change right so yeah yeah it was a it was a beautiful story I’m I’m super glad for

super happy that viim gave me this opportunity I it’s it’s amazing uh yeah

it was it was awesome good good yeah no thanks for thanks for sharing uh that

that’s that’s great to great to hear makes me makes me smile well um all right so what’s the application of the

SC age I I think people are most interested in in like the application based questions because they just want

Applications and implications of single-cell clocks in aging research

to know like what does it do how do I change how do I get healthier you know what what is how does this benefit me um

is what people are thinking yeah so I think I think there’s a lot of

applications and it it really opens up like a whole new kind of uh area to to look into so I think some

of the applications that we explored in our paper are really just the beginning and I hope that that people kind of look

into a lot more things but one of them is like uh so we first looked at these

hpes which are the most predominant cell type in the liver these produce albumen and they’re kind of crucial cell types

and when we applied the tool to to young and old heid you saw that they they aged right so the old ones were predicted to

be younger than the young ones which was quite cool because for some time people thought that there was a huge bias of

cell composition in the clocks and I think there still is to some extent unless you explicitly correct for it and

there’s some really great work uh recently by Al Alan Tusc at the buck looking at some of these uh

differentiation independent methylation patterns and harnessing those to make H clocks that are independent of cell

composition but um yeah I think like it it was for

the first time like some evidence that individual cells themselves age because he apdes the older ones had a different

methylation profile than the younger ones and judging by these age related Trends um there there seem to be

something there another interesting application was this muscle stem cell data so it looks like at least

epigenetically on an aging Spectrum the muscle stem cells don’t age very much and this had been already shown by wolf

in in his original paper uh just not at the Single Cell level so we just kind of extended that to the Single Cell uh we

looked at embrionic stem cells too and it looks like depending on how you culture them there might be some differences too their biological age

although again this is a very nebulous term and I think people should not over index on on kind of these metrics they

are very important and I think they will perhaps revolutionize kind of the longevity space but there still needs to

be still needs to be a lot of work done in there for sure but I think the most interesting one is kind of the last two

figures of the paper which we’re looking at embryogenesis and so just a couple weeks before I was starting to work on

my paper uh my colleague chabba who I mentioned he was doing some really cool work on applying conventional epigenetic

clocks to embryogenesis data so this is the period after fertilization all the way up to basically birth but really the

first couple weeks after fertilization and there was a a theory that viim had kind of put out there in a

in a an opinion piece that perhaps cells

rejuvenate during this period of embryogenesis Because if you think about it um when people have children the the

gametes so the the sperm cells and the oyes they’re old right they’ve aged with the person maybe they’ve aged less CU

they’re a little bit less metabolically active but they’ve still aged because they’re in the body they’ve been there um the stem cells are aging etc etc so

somehow they age but they need to give rise to an embryo every time that is of age zero otherwise every generation

every subsequent generation would get older which doesn’t make any sense and we’re not seeing that right um and so

kind of mapping when that Rejuvenation process happens during embryogenesis is a really interesting task

and Chaba did some really good work on that doing it all at the bulk level but for some of these like really early

embryogenesis time points there are just so few cells that it doesn’t really make much sense to do it in bulk and so one

thing that we tried was using one of Wolf’s data sets and this is a fantastic nature paper that he had published in

2019 we applied the method and we saw very similar Trends so basically from

embryonic day 4.5 so 4.5 days after basically the formation of um of the

zygote we see kind of from 4.5 to 7.5 we see a really nice decrease in epigenetic

age and I think that was another kind of hint that indeed maybe this Rejuvenation period does does occur um in mice and

probably in humans as well because otherwise it be be difficult to explain how we’re able to reset age essentially

and I think it’s like super pertinent specifically when people talk about like reprogramming and Rejuvenation because

our body naturally does this like we’re trying to use some of these artificial methods meing the yamanaka factors the

okm are a great start but they they were they were artificially discovered right to turn fiber blast into ipscs whereas

here our body like naturally kind of has this reset of the entire epig genome and

somehow that seems to be correlated with the decrease in epigenetic age using clocks that are only trained on normal

aging right so not at all in any embryonic data so you just apply kind of out of this out of distribution and you

still see these Trends and I think several people have kind of substantiated these results now so I

think it’s just very interesting to to kind of explore that and I think single cell is is a great way to look at those

things and but beyond that lots of applications I mean you can think about maybe Discerning if some cell

types age at different rates than other cell types if some cells in a distribution are older than other cells

uh if you give an intervention to some cells maybe some of them get younger maybe some of them don’t understanding

why that’s the case I mean a lot of it opens up a whole Pandora’s box of just kind of a lot of applications yes let’s

could go on and on like I said it’s it’s like a field within itself I think the single cells and well I mean the immune

cell deconvolution and you know all the work they’re doing at the buck Institute um yeah goes hand inand with this a

little bit right really identifying down to the cellular level why are these markers different how do we go through

that reprogramming process so just so many questions um and that you all are you know working on and and chewing off

and trying to get get the answer too so we’ll we’ll see what comes from it all

indeed um are I I know kind of when we were talking about bulk um measurement

um and and then the Single Cell measurement you talked about some of those limitations already but would you say that there are any further limitations with your um SC age that you

you would want to improve on yeah for sure I think you can always

you can always improve on these methods I mean one thing would be just getting better data right so uh I trained all of

my like like reference models for aging based on bulk of like a whole tissue so

it was like bulk measurements of liver across different time points in mice ideally if you only want to measure age

in hepatocytes it would be ideal to just have hepy profiles and maybe you’re still doing these in bulk but you’ve

like sorted just hepatocytes right so instead of like I mean the liver is fairly homogeneous to mostly these seyes

but there are also kufer cells and endothelial cells and so the composition of those might be changing and I think

some some work from the tumorous Consortium indicates that there might be kind some change with ag so to abstract

away fully from that Exel composition and to stop it from biasing your results I think it it would be useful to be able

to train those but just again the data just doesn’t exist but I’m very hopeful that in the next couple years kind of people will will start to make that kind

of data of like cell type specific methylation profiles either in bulk that they can be used uh for single cell

ideally you would even want to train a single cell model to apply to single cells but again the spicity constraints

at the moment are a problem but there’s new methods coming out uh like RBS which is reduced representation by selfa

sequencing which is a more targeted version and you can even do even more targeted than that so you can restrict it the time seek paper from Patrick

Griffin does a really great job of this of using probes to restrict basically your measurements to only specific age

related sites so if you’re able to consistently measure those you could use more of these conventional approaches

that people have tried in the past on single assell data uh which would be which would be great and I think again

there are also probably algorithmic improvements that can be made in feature s and some of these have been have been tried by by Wolf’s and his colleagu in

in his most recent preprints but yeah there might be some some additional things that we can do for sure you can

always improve on these methods I mean if you look at single solar in uh really started about a decade ago and just the

absurd amount of methods that have come out now and they each claim to be better than the previous not always not always

the case but there’s yeah the tremendous amount of progress both like on the data acquisition s so just experimentally but

also on computational side and I think the algorithm that I used is really simple so there might be much more

complex ways to get a better better prediction for each individual so yeah help reduce the noise of the algorithm

and and kind of TI tie some things in there um no that all makes sense um well

yeah what about the future of epigenetics in general do you what

what’s your you know your thoughts your opinion there do you think and I I’m a firm believer and everything has an

application the first generation clocks were created for a reason right when we talk about the bulk tissue ones yes

they’re trying to predict chronological age but they were used to help you know date refugees to see if they were old enough to seek Asylum or for Crime Scene

investigations from you know a DNA sample so I think everything has an application in a use but I guess what

I’m trying to ask you is if if you think the field of epigenetics in the future is going to be built more on the Single

Cell clocks or or more of the bulk tissue methods um again depending I guess on what you want to look

at yeah yeah that’s that’s a tough question I mean I think again there’s value to both both methods I

think the more we build kind of these the more we study heterogeneous systems

and the more we build uh kind of these perturbation strings or things like that that impact different cells in different

ways so like one cell might get one genetic perturbation and another cell might get another in those cases you

really need single cell because you’re you’re not really going to understand anything from just averaging um all the

cells together and I think this is a great time to bring up this this classic analogy of the smoothie right so with

bulk you’re looking at the smoothie of cells if if every cell is a fruit you’re looking at smoothie and that’s what most

clocks do and you can tell kind of a young smoothie from an old smoothie but if you’re not understanding how the

individual blackberries and blueberries and strawberries are then you’re missing a huge amount of information and uh I

think yeah like most of biology Now is really moving pretty heavily into single Sol approaches again they are still huge

constraints to them they’re still very expensive for many LS although the cost is coming down every year but it’s still very expensive to run uh quite a bit

more technically difficult than than bulk as well and then the analysis too a lot of batch effects a lot of things you

have to correct for but I I’m hoping that kind of things move more and more toward single cell and I think we’re

starting to see also some of the first clinical applications of single cell specifically kind of in these high

heterogeneity environments understanding tumors and immuno oncology and things like that I think that’s really where

the promise of single cell lies and you you can do this across multiple modalities and gentics I think is is a

huge one as well so these bull clocks will continue to be around I think they’ll be they’ll be cheaper they’ll be

like the time seek paper for example can bring down the cost to just a couple dollars which I think is really incredible from from the current like

several hundred dollars that you need to pay for for one of these methylation arrays that’s in bulk but if you could

have a version of single cell I think single cell is going to be much more of a research application than it is

um yeah I think maybe on a consumer level it would be much easier to do kind of these bulk epigenetic age

measurements for a long time but on on a purely research based I think understanding how individual saws

respond to things and how yeah they differ is is also just yeah really

interesting yeah I agree with you on that love the Smoothie example by the way that’s great um I’m GNA make that a

highlight so people can understand um you know when we’re really digging into this so no that’s a really great example

and and I agree with you um you know commercially there’s all these biological age clocks and

commercially even like the biohacker consumer people who

you know are are probably more well versed in the space than the the the general consumer um still may not

understand the differences between all these epigenetic clocks and what’s going on right so so I think it’s going to be um harder to introduce the Single Cell

to Consumers um but I do think it is the future especially as we’re we’re trying to understand causation rather than just

correlation and what’s happening at the cellular level so um yeah I mean it’ll get cheaper as more

people are using this technology and hopefully converting to that once there are more insights those Interventional trials and things so um I just I think

it’s super fascinating totally yeah in regard to this causal argument that you make as well there’s a great preprint

from one of my colleagues in fim’s lab Albert who kind of starts to explore some of these these causal relations

which are still totally kind of undetermined right like is the damage in

the methylome what causes Aging in some way or is it a consequence of Aging or this whole like damage hypothesis like

how does that link to these epigenetic clocks we see very concered changes in particular cpg sites with age so is is

some of it stochastic is some of it not stochastic and Andre also from BS lab who’s now retro has explored some of

these questions so I think there Still Remains a huge amount of mystery with how kind of aging and epigenetics mingle

but at the same time it’s it’s really cool to see a lot of great work in that space definitely yeah exciting

nonetheless um but yeah for familiar with those those papers and and what you’re talking about there um so let’s

Future directions in epigenetic research

focus on you what are you doing now what is what is retro you know bioscience tell tell us

more I know there’s been some things in in the press and you know I’ve been trying to to keep myself updated but um

would love to hear about everything you’re doing yeah it’s uh it’s an amazing place for sure um been really enjoying my time

here I joined really early uh so a scientific team with just a handful of people and I’ve been building up the

computational team since then uh we’re basically eight people now which is uh which is really cool uh I think we just

do du and very complimentary backgrounds as well so I think we’re going to be able to accomplish really awesome things

the whole company as a whole it’s about 35 people uh were based in Redwood City in California and I think it just came

out last week that Sam Alman is our investor as well so he’s the fame CEO of

open Ai and and previously president of white combinator the most legendary uh kind incubator for startups uh and it’s

really great to to be able to do I think one of the things that really defines retro is is well there there are many

things uh first of all everyone’s like super mission mission aligned and Mission oriented I think we’re all just

really interested in the problem of aging and how we can at the root kind of solve that um as opposed to just like

going for like a particular age related indication right like how can we

understand aging itself and ideally make therapies that that that work against staging uh we’re also really hungry so

we really really want to do this uh very kind of hacker like Scrappy culture as well um we we have a warehouse full of

containers which is um kind of really unusual for a biotech we’re not kind of like the classic um biotech like fancy

lab with a bunch of windows it’s like these like really funky like futuristic containers but I think it’s a great Vibe

I think it’s it’s really awesome and that was that was Joe’s decision back in the day yeah it’s uh it’s a really cool

place we’re culture is what what makes it so yeah I think so too I mean it’s it’s ultimately What drew me here I had

interest um from from many of these Ro programming startups and and kind of longevity companies when I was U

planning to make my exit from Academia and when I met the people at retro I just knew this is the this was the place

that I wanted to be at so yeah we’re interested scientifically in a bunch of things I think reprogramming is definitely one of the big ones and and

also to be able to do all the science in a fairly financially unrestricted sense is really great like I think in Labs

I’ve been fortunate to always have been in quite well funded Labs but I know a lot of people who have a hard time kind

of getting the experiments through because that they like funding or they need to write grants so just having a huge amount of cash in our case $180

million to to explore this is really is really clutch um and it just enables you

to kind of be very creative and if you have an experiment that you think would produce meaningful aners you have the

backing to do it and even if it costs several hundred thousand uh but more scientifically

speaking we’re really interested again in reprogramming across a couple different cell types uh we definitely have a quite prominent T Cell program so

we’re interested in rejuvenating t- cells which are a big part of the immune system and they’re involved also in in

not only vaccine responses and like regular responses to to disease but also immunotherapy I think is a huge one if

you can improve immunotherapy by somehow rejuvenating te- cells that would be super cool one of the things we’re

interested in we’re also interested in other cell types so we’re exploring we have programs in a couple other cell types as well and then I think

reprogramming while it has the most potential in my opinion for really like resetting age and rejuvenating cells

should also not be the only thing that people look at so we’re also really interested in how you can modulate the autophagy process in cells and how that

can maybe be um uh either rejuvenating or somehow delay or slow down aging in

some way and then also these plasma Therapeutics which is related to some of the parabiosis work that was kind of

pioneered by by Arina Convoy and Tom rando’s lab who now has her own lab in Berkeley and also um some work that I

did um who which actually just as of today is impressed in nature aging so looking forward to sharing that with

people in the future but um kind of how parabiosis which is a really gruesome

procedure when you link an old mouse and a young Mouse and Ling the blood system for for quite some time people had done

it for uh short amounts of time but we were the the first in Boston to do it for three months in collaboration with

Jim White’s Lab at Duke um and then that showed some like rejuvenative um

phenotypes especially on these clocks it looks like a lot of kind of the Hallmarks Hallmarks of Aging were were

diminished as a result of being exposed to Young Blood which is obviously not super practical for human applications

you can’t just connect a young person and an old person I mean there’s there’s a great episode of Silicon Valley that

is very much about that right like these like blood boys um yeah but yeah I mean

if you if you can figure out what specifically about young blood is good like what what maybe youthful factors

you can take or manufacture in some way to give to older people or vice versa what you can take from older people like

some of these negative like old factors in Blood and somehow remove them all through kind of modulating the plasma

the plasma parium I think that is like much more proximal to the clinic and so

that’s also a really interesting kind of space to operate in because it looks like this dilution hypothesis has some

Merit it’s been a couple couple recent papers that shows that if can dilute um

if you can yeah if you can dilute plasma um you you may uh improve a variety of

age related facets and so kind of understanding what proteins are are important there is kind of the Crux of

the equation and so that’s that’s one of the things that we’re interested in as well yeah amazing I I yeah my I mean I’m

just wow and and yeah just so excited about everything that you all are doing

um I don’t I don’t know what I’m most excited about probably yeah I think the

you know what I spend most of my time on is you know talking to healthcare providers about this this epigenetic

based testing in the integrative functional medicine space and a lot of them are doing like the young plasma f

um exchanges or um uh you know working with patients on

on yeah figuring out what factors they can include for for medication to try and make those clocks uh tick tick in uh

hopefully backwards in the right direction lowering your your biological aging processes so I think the application of that one is really

exciting I think it’ll be as um prevalent as you know like IV clinics are now like you’ll see them on every

corner in like 10 years or something um and you could just like text your friend hey what’s up want to go get like young

young plasma exchanges and it’ll be a thing um so I think I think that’s super

exciting as well um but yeah Alex it was it was great to learn you know about

your single cell work um the clock that you created even what you’re doing at ret is there anything else that you want our our listeners to to hear or um you

Alex’s work at Retro Biosciences and its potential impact

know if if people want to connect with you where where they can uh find you as well yeah for sure um so first of all

retro is hiring across multiple positions so if people are interested um computational team is is a little bit

full right now but we’ll be hiring again probably this year but all the other teams all the experimental teams are hiring so if you’re a scientist and

you’re interested in longevity and you really want to make an impact um yeah definitely can’t recommend retro enough

I think it’s it’s one of the coolest places I’ve ever been in um look forward to being here for quite some time as

well uh yeah and I I think um it’s been really awesome to see more and more

interest in the longevity space recently I think like the formation of some of these fellowships like the work that

Nathan Chang has done at odl and now now with with Mark also at the uh longevity

biotech Fellowship so if people are interested in that I’d recommend they check uh they check it out it’s just

it’s a way for people who come from non-traditional backgrounds people who maybe haven’t studied biology to try to get into longevity I think it’s it’s a

really awesome effort as well um yeah more Community more awareness uh it’s been awesome to see also a lot of

funding come into the space um but also realizing that these things take time and they’re really hard to do right so I

think there’s been a huge amount of hype with reprogramming with you know of alos and and UL limit and

retro but these things take time and clinical trials and things like that will take time as well but I’m

optimistic that well be able to to really have quite a powerful impact on

uh yeah yeah on on longevity and on aging and and ultimately retr mission right to increase healthy human lifespan

by 10 years and that’s a super difficult Mission very ambitious right because yeah maybe like antibiotics and hygiene

has has drastically increased lifespan in the last like several hundred years but like to get to the point where

someone is already healthy like already exercising doing all the right things for their longevity and then giving them

an additional 10 years of healthy lifespan right not lifespan where they’re um where they’re disabled or

where um they really can’t they can’t do much um where they’re like that stricken right I think like having 10 more years

is is is super cool and even if you can just get to one or two or three I think it’d be it’d be really awesome because

like as yeah I think we’ll start to see that soon is is what I was gonna say um I I

don’t think it’s I unheard of I think it’s like super reasonable I mean it’s not going to be easy right um but like

you said with the team you all have and you’re so Mission driven um it’s it’s the motivation and the why behind you

know all of your your work that I think is just incredible for sure and science is unpredictable too right so you never

know really what what will happen and but I think it’s like fascinating research questions to work in and and

hopefully we’re able to translate them into the clinic within a reasonable amount of time and I think that would be

the most fulfilling right if you can see a therapy that you were you were involved in creating and it provide

someone with like more time to spend with their family or more time to do some of their favorite hobbies or yeah I

think it’s it’s really cool and I think also one one last thing is I think like the the advances in artificial

intelligence are also really really interesting and the the uh the the kind of mingle between that and Longevity is

potentially very powerful I think feel like some of the recent um recent announcements from open AI for example

gp4 this week and chat GPT couple couple months ago I think if you’re able to kind of harness some of these like large

language models and also just like you algorithmic improvements that have occurred in the last couple years in the

field of deep learning and artificial intelligence you could really make quite fundamental dense into how you

understand biology in the first place I think right now all our models and all our representations are pretty good but

they’re not perfect by any means and I think it’ll be hard to get anything perfect I think biology is like very

noisy and finding the signal in the first place is really difficult but it’s a really exciting time to where you can

maybe leverage some of these advances in Ai and merge them with Biology and also both from the experimental side improve

methods and also on the computational side to just understand things better and then ultimately enable uh enable

more of these uh these therapies to come to light I think would be super cool yeah I think so too I think that

was well put that’s just like a little bow wrapped on our on our entire conversation um so one last question for

you Alex I ask everyone this at the end of the podcast it’s just really fun question um if you could be any animal

in the world what would you be and why oh wow I wasn’t ready for this one um that’s a really interesting make

you think I used I used to ride horses when I was younger so I definitely definitely have a good vibe with a horse um okay I

also really like cats and dogs I had a dog for for 13 years golden retriever um

oh love goldens I grew up with goldens yeah yeah they’re great dogs would see

H I like owls too I think owls have a good vibe owls are cool um that’s unique

yeah no one has said an Al yet they’re just kind of like chilling in the trees and they’re kind of nocturnal as well so

I recently saw a llama too in California it’s unexpected site to see a llama but

yeah I think um yeah I like to think about things so maybe maybe an owl would be would be a good one perfect yeah

we’ll we’ll um we’ll we’ll put it on the table but not not confirmed yet so maybe no confirmation indeed yeah and if

people also yeah sorry just if people want to to reach out to me as well um yeah I’m on Twitter um Alexander tra is

my my handle you can find me there you can also just email me Alex retro. bio

um feel free to reach out if you have any interests or questions or especially if you’re interested in what we’re doing here I think it’d be be cool to to

chat yeah yeah just to to open up the conversation well I’ll put everything in the show notes um if anyone has any

questions for Alex send him to me if if you can’t find him we’ll we’ll we’ll make sure that they they get to him but

uh yeah thank you so much everyone for joining us at everything epigenetics podcast remember you have control over

your epigenetics so tune in next time to learn more thanks so much Alex yeah thank you it was a pleasure