Development and Application of Epigenetic Predictors

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

your health I’m Hannah went and I’m the founder of everything epig genetics

today my guest is Dr Veron dwarka I work

with Veron as he’s the head of bioinformatics at true Diagnostic and I’ve known him for about three three and

a half years now and we have became good friends over that time this is a really

cool conversation because I’m not always up to date on all of the bioinformatics happenings at my company so we really

dive into his background and how he got involved in in epigenetics and what he

really studied during his graduate school years but we further get into you

know what’s an epigenetic clock I previously discussed the generations of epigenetic clocks on my podcast but

really how do we make these clocks what are the best bioinformatic practices when we’re creating these types of

clocks and we talk about reproducibility using principal component analyses

corrected algorithms making sure these clocks are related to clinical outcomes and have Association to disease and we

even talked the validation of these clocks and what that looks like and that it’s important to use the specific

tissue type the clock is built in my favorite part of the discussion dives into these new EBP or this idea of

epigenetic biomarker proxies and we actually talk about the difference between an EBP and an mrss or a

methylation risk score we end with what other areas of Discovery really excite

Veron in the specific projects that he’s working on at the moment a little bit of background

information for my guest like I mentioned he is the head of bioinformatics at true diagnostic an

aging and Longevity investigator that specializes in epigenetics and

bioinformatics he’s co-authored numerous Publications relating to genetics epigenetic DNA methylation and tissue

regeneration he currently serves as a 2023 foresight fellow in biotechnology

and health expansion awarded by the foresight Institute Dr dwarka is

passionate about implementing machine learning methods to advance predictive medicine identify novel biomarkers and

create algorithms to better understand the biology of aene now for my guest run

darara welcome to the everything epigenetics podcast maroon I’m excited

to chat with you today yeah this is awesome and by the way can I pay you a compliment before we start because

because this podcast has actually been one of those podcasts I’ve been listening to um mainly because I commend

you for how you’re distilling a lot of the epigenetics world into something that is approachable and so hats off to

you just thanks Veron I appreciate it kicking it off with with a compliment

well cool no I’m super excited for this episode as I alluded to in the

introduction Brun and I work together um we first met I don’t know maybe what two

or three years ago now which seems crazy it’s coming on to three yeah yeah about three years ago yeah so I’m I’m excited

I’m excited to dig a little bit deeper you know we’re we’re good friends as as well we get to see each other in in

office every day and work on some really cool stuff so you know when we first met

and uh it it was yeah like I said three years ago now you lead the bioinformatics team at true diagnostic I

think you are the most knowledgeable when it comes to biomarker Discovery using machine

learning AI really large data sets through the lens of epigenetics and

creating all these different predictors of of Health using that data and we’ll we’ll talk into exactly what that means

here during the rest of the podcast um but tell us a little bit more about how you you got here and how you ended up in

Lexington Kentucky too sure yeah well thank you first of all for mentioning that um I yeah so I I guess like

starting off I grew up in California in the San Francisco Bay area and so like when I did a lot of when I wanted to

kind of pursue um what I wanted to do in the future I kind of naturally gravitated a little bit towards uh

biology even though I was a musician growing up and that’s kind of what I wanted to do um and how I ended up in

Kentucky is that actually my undergrad uh institution I actually went to University of California Santa Cruz and

that actually started this whole bit on bioinformatics so I didn’t start in epigenetics I wanted to understand how

we could utilize computation and Technology to better understand biological Big Data essentially the all

the data that’s coming out of these sequencers these A’s G’s C’s and T’s and also the marks that are around it um how

does that relate Back To Human Health not only just human health just biology in general and so um when I went to UCC

Santa Cruz um I worked at the UCSC genome browser got to know a little bit more of that and so when I got to

Kentucky I specifically sought out a professor uh Dr Randall Voss at the department of biology to work on

salamanders on Axel of all of all things because they had this in insane ability

to repair and regenerate whole tissues and structures and so I studied TI

tissue regeneration and one of the things I focused on was how are the genes being activated and also

deactivated during this process um and one of the things is when you start to look at you know genes in

terms of a circuit board you know you know you can kind of understand okay these genes are being turned on so the

the switches are on these genes are turned off so these switches are off I started to navigate towards okay well

what’s causing that switch and it kind of ended up on epigenetics and so I

actually a large portion of my of my final chapter um utilizing a lot of um

you know uh DNA methylation specifically DNA methylation techniques to kind of

address those questions and when what year were you in uh Rand boss’s lab when

was that it was uh 2015 to 2021 okay so that’s so weird because I’m

actually just putting this together for the first time I was at UK from 2015 to 2019 and like I remember thinking of

looking into Dr ROV vos’s lab because he had the cute little salamanders and the Axel LOE and being able to study those

for tissue regeneration like I remember going and trying to find a pi in someone’s lab that I could be a part of

as an undergrad so like I was in that same biology chemistry department and

like I’m sure our path crossed but I’ve actually like never thought that out loud so that’s interesting fun fact yeah

we were in the same department at at University of Kentucky um when I was a grad student I think you were a um

undergrad at under yeah yeah yeah well so super cool and uh like I said I’m

sure we we cross paths if not then you know eventually at at some point which of course we did at at true Diagnostic

and you know before true diagnostic I didn’t really know much of anything about Fe genetics I get that all the

time like how do you have all this knowledge and it’s it makes me giggle because it’s really all self-taught like

I have really good teachers like you and everyone else in in the company and all of the academic and institutional

partners that we work with so digging into this literature and um again learning from my surroundings is is

really the way we’ve been grinding at at true D but you mentioned you know previously that a significant portion of

your dissertation covered epigenetics specifically looking at that histone U methylation changes in DNA methylation

changes in salamander tissue regeneration so not necessarily human data but there’s definitely going to be

you know that overlap and what I’m really getting at here is you know epigenetics has exploded dramatically

over the past decade even over just uh the past couple of years so you know why

do you really believe this field has been rapidly exploding and and like why do you think studying epigenetics and

these changes is is important broadly speaking broad terms I think yeah no I I

think it’s because of um that same experience that I felt because um I think in the later part of let’s say the

2010s a lot of people were doing um like RNA sequencing so looking really at how

genes are being expressed and uh de expressed and how these circuits are being there I think that you know a lot

of work is still continuing on but I think we’re point at this point of saturation where okay we know that in

you know make you know just figure out some kind of disease we know that these genes are expressed in that disease the

natural question really comes on to what is Upstream of that what is contributing to that overall you know gene expression

pattern and if if uh you know you kind of go back into bi biology 101 there’s

this whole concept that I always talk about called central dogma where you know uh DNA turns into RNA RNA turns

into protein you have transcription translation and epigenetics um kind of

serves as the Upstream of that transcription so you know when we talk

about epigenetics I have to say you we not only just discussing DNA methylation but we’re also mentioning non-coding

rnas histone modifications and all that but DNA methylation kind of comes into play because that is um that that is one

of those um marks that is readily associated with a lot of um

environmental factors and so when you’re starting to understand in terms of human

disease what that Upstream Factor could be there seems to be a lot of literature there actually is not seems there is a

lot of literature that suggest that DNA methylation might be the specific epigenetic Mark that really you know

kind of goes Downstream into the uh central dogma and so that was I think that’s why epigenetics and spe

particularly DNA methylation in the context of human health has really exploded yeah definitely and I’m

obviously not going to throw them out there or you know mention any names but I think to just as researchers in this

field we’re seeing a lot of people in these really large Labs who are betting

on transcriptomics previously now betting on epigenomics and you know quite

literally transferring all of their funding all of their sources to now studying kind of a different omic layer

so we’re seeing epigenetics like I mentioned just explod uh and and we continue to I have no doubt in my mind I

mean there’s you know a new company created every single day a new lab created every single day that’s actually

trying to understand what these insights mean or even epigenetic reprogramming and trying to set these ages of of cells

so to speak back to zero to figure out how we can do that on a larger scale yeah agreed agreed and and I think

that’s what’s the beauty of it all it’s that you know no one stumbles into this you know I I stumbled in onto like as as

an accident I I vividly remember being in my advisor’s um room uh office room

which was actually it’s funny he he used to work with the salamanders like he had

his own office nice little office but he would always stay in the stock Center um I remember coming into there and I was

like look you know I naturally like what is causing all of these switches to be the way that they are

and he was like well why don’t you find out start with epigenetics and he literally gave me a data set and was

like just learn it there is a textbook I have on epigenetics um I maybe looked at a few P

pages but a lot of the learning and I think that you feel the same way people in our company feel the same way and even other people in the field feel the

same way you have to go to the primary literature because this is such a it’s not a nent field because Adrien bird

back I think in 1967 60s and 50s really did spearheaded a lot of the

understanding of these DNA methylation marks but um I think moving forward a

lot of I think the boom of nextg sequencing actually has you know made it a super exciting and um field which is

now rapidly evolving so I think the primary literature is really where you’re going to learn a lot more of it um in the context of today definitely

and I would not know this answer off the top of my head so no pressure If you don’t know it Bru do you know who wrote that epigenetics textbook by chance

I believe it is I have it I think it’s like C Alice um I think it’s Alice last

name is Alice interesting okay yeah well I’m just like interested to see like how

long ago it was written and like what’s included in that I yeah we we need to write a book on on epigenetics it’s like

I don’t even know where you would start um but it would be like so introductory level I feel like um yeah and you could

go into you know any in all different areas of Health but um like many of you know

at my company true diagnostic we really use epigenetics well started out using epigenetics to measure aging because

that’s where we can actually have the largest impact so this is a really good segue into what I want to talk to you

about next Veron you know aging’s that number one risk factor for all cause mortality morbid morbidity so not only

is there a massive Health Care impact that we’re talking about here there’s a huge financial aspect too and for those

um some some people listening may have seen Dr David Sinclair’s paper he wrote on this um in which the economic value

of targeting aging is just huge so for example if we increase life expectancy

just by one year it’s worth $ 38 trillion and the US debt is around $34

trillion right and it keeps growing and growing so this is usually a number this

biological aging I’m talking about here in like a clock format and I don’t know that I’ve ever actually

you know even Define what an epigenetic clock is or you know more with your skill set Baron on how these are

actually created and made and you know what really defines a clock so can you even just in in simple terms explain

like what’s an epigenetic clock or what what is a clock what is that mean sure yeah a clock is kind of derived from

this idea of a biological clock and so um based on your date of birth uh the

time that you’re born you can have something called a chronological age clock where it’s based on the time you

were born and the date that you were assigned now essentially or that you have now a biological clock is

essentially the age in which your body represents so you might be 24 but you’re

you have the body of a 22y old or you’re 34 and you have a body of a 24 year old because you take really good care of

yourself this goes back into how are you estimating that biological age you can estimate biological age multiple ways

you could use peptides you could use proteins you could use uh clinical markers um and so an epigenetic clock is

utilizing epigenetic markers specifically in this case we’re talking about DNA methylation but I will have to

say there are transcriptomic clocks and um histone modification based clocks as well but in when we talk about

epigenetic clocks largely were’re focusing on DNA methylation markers it’s essentially the

orientation or the methylation of the DNA methylation sites contributing to

that biological age calculation from a number perspective essentially so an epigenetic age clock is your biological

Age based on your DNA methylation essentially yeah and there’s different types of those as well right so I think

it’s essential to even just Define a clock like you said you can create a clock with many different omix or many

different biomarkers these epigenetic clocks of course they going to be using D methylation data or epigenetic data

and I think I have previously explained the difference between the first second and third Generations right first

generation is just using chronological age information along with DNA methylation second generation is going

to be better because it’s using not only the DNA methylation but other underlying biomarkers so it’s going to be a better

predictor and then when we think third generation remember we think of that denan pace algorithm more of this

longitudinal analyses so and we love that dunan pace algorithm yes yes we love the dunan pace algorithm it’s very

unique and it’s still my favorite one to talk about to this day um and this is

why I wanted to have you on ver with your expertise in bioinformatics you know statistics um what are the best

bioinformatic practices when you’re actually creating these these types of clocks and and feel free to really go in

into detail here as well absolutely um I think based on the based on the

experience that we we’ve had um especially a true Diagnostic and also just know vetting a lot of these clocks

um the first one I can think of is the reproducibility so I think there’s a

there was a paper by Dr Higgins Chen um with Morgan LaVine when she was in Yale

at Yale University where they actually looked at how a lot of the clocks that were are currently out there so those

first generation second generation third generation clocks and what they started to do was or notice is that you know the

reproducibility that like you take the same biological sample and you you know quantify the methylation the variability

in those uh in those outputs weren’t that great you know You’ have like SKS

of like 3 to four years difference from the same exact output and so when we so

they started to utilize this metric called icc’s interclass correlations which is essentially a correlation

between how reproducible or how correlated are the same values um from repeat tests and so when we start to

address the first thing in to address these clocks is how reproducible and how

correlated are those repeat measures and one of the reasons why why I we always tout um do need in Pace and additional

clocks as well you know I think the reason why I brought up uh Dr Higgins Chen is because they actually um

recreated some of those first second and third gen sorry first and second generation clocks using this principal

component technique Main and the main reason was to improve this ICC this

intraclass correlation or reproducibility of these epigenetic clocks so that’s number one and can you

just very briefly to Define what principal component analysis means yeah yeah yeah yeah so principal components

are it’s a statistical measure method where rather than looking at individual

cpgs you can uh you can Clump multiple cpgs or group multiple cpgs that are

numerically very similar and so what this allows is reliability now when

you’re doing a lot of these non-pc clocks or principal component clocks you’re estimating individual cpgs that’s

great but the problem is actually when you run let’s say you run a sample

through you know this methylation quantification if there’s some technical error in one cpg by random chance that

is highly important for the quantification of that clock then you skew the entire data set right that

value doesn’t actually work out too well and so why this principal component was a really actually really neat method is

because now you’re not privy to just one cpg because even if that one cpg fails

all of his buddies like are still there all of the other PCS or sorry all of the other cpgs that are part of that one cpg

are still hopefully you know if unless if it’s some major issue are still relevant and they are still you know

being Quantified in a very um positive way and so that allows for that reproducibility to actually occur I like

that I it’s the princip pal like all your buddies your little cpgs I’ll start describing it that way and when Brun

saying cpg he just means that DNA methylation marker that cytosine phosphate and guanine if we want to

really start and Define some of these terms um but that’s a good point I I usually explain it to people as well

isn’t it when when you have the principal component analyses algorithms isn’t that like thousands and thousands

of of cpgs included in those actual algorithms that’s precis correct yeah yeah a lot of these non-pc or principal

component based clocks will only U maybe measure about 350 let’s say three

according to the horbath clock but when you start to look into principal components you can actually incorporate

M many more cpgs just based on how the methodology kind of works out so you’re

actually imparting more of the signal yeah they’re kind of like leaning on each other I I almost think of it as I just think of it as like adding another

layer of like QA QC almost absolutely I don’t know it’s the kind of most simple

way i’ I you know felt like I could describe it to to help that I think you are adding an additional qaqc to it so I

I don’t think you’re off base there yeah all right what else so you have you know reproducibility adding in these

principal component analyses clock which you know when you’re adding those on top of the Baseline algorithm it accounts

for a lot of those differences so that’s really important um maybe the next yeah top two or three best the next thing

that I could think of is honestly um like you know again if you want something that’s clinically reliable you

want it you want to also know like what its Association is to clinical outcome and so um this is I think where it

really separates out um some of the first generation clocks to let’s say the second or third generation clocks is the

association to disease um I will say this the first generation clocks actually have associations to diseases

right like the um the Horvath clock actually shows significant Association to cancer risk and so depending on what

the question is or what the clinical outcome that you’re really interested in will dictate which of these clocks are

are there however once you start to incorporate more of the biological let’s say signals like for

example from uh pheno age which actually is trained not to chronological age like

these first generation clocks but more to a phenotypic score that is based on

biological markers what you’re actually doing is imparting a lot more biological input so you’re you’re not just basing

it again on something that is you know constructed by chronological age you’re constructing it more on the biology and

what you start to notice is that when you let’s say associate to time till death let’s say you want this measure to

reflect how long a person might have to live mortality what you start to notice is

that these biologically derived clocks show stronger associations to overall

outcome mortality outcome than let’s say these chronologically uh trained algorithms and this is why that

Narrative of of the second generation being better kind of comes into play is first generation does its job and it’s

still a relatively you know it’s a clinical output that’s relatively um relatively positive but when we want to

look increase the breadth of that clinical utilization that’s where the second and third generation really come into play and I will say one thing too

is that the third generation I know that there’s a there’s a u kind of a differentiation some people call the dun

Pace the third generation and the and there’s like a Mamon clock that is now third generation the that I’m using it

third generation I’m representing Dyan pace so just wanted to make sure that you know the listeners are clear about

that got it yeah the language is important here because there’s a lot of unknown terms or you know terms that have been defined in multiple or yeah in

a multitude of ways so yeah you know you’re right first generation I like how you said it does its job second

generation yeah they’re probably better right well we know they’re better because they are more predictive of of

outcomes of all cause mortality and morbidity so that means if you’re test ing a second generation clock and your

biological age you get back is actually older than your chronological age um you are at more increased risk for all cause

mortality morbidity it’s going to be more predictive and same is true of the opposite if you test in your biological

ages younger than your chronological age great even better you’re at even much

more of a reduced risk of all cause mortality morbidity and actually in it it’s it’s it’s funny cuz today um just

today we we put out a paper in nature aging which was talking about stochastic

clocks now the stochasticity of it all is I think this is a term that might be

M like it was hard for me to understand initially when we were kind of just even coming like talking about the study but

this idea is that there are two aspects of Aging that I think people need to

consider what is your stochastic aging and what is your Dynamic aging your stochastic aging is the Aging that

you’re acing throughout your life so this is you know as you exist in your environment you are garnering changes in

your biology even from the molecular side in terms of DNA methylation you’re accumulating DNA methylation as you as

you just live EXA and so that’s you’re going to be your stochastic aging these first generation clocks do really well

in capturing that stochastic agent and that’s what this paper um kind of talks about however there’s also a non quasi

stochastic or more Dynamic um aging that’s also being captured and what we

show is that you know these biologically derived clocks namely uh pheno age

actually does a really good job in capturing that Dynamic uh aging which is more Associated to you know in a in a

short amount of time that you might have a rapid increase or decrease in that age which is might be Associated to disease

onset or let’s say improvements in overall um health health um Health

outcomes and so that paper was really um we were really thankful to work with uh

Dr tessendorf and his group and also Dr Jess gasis Su but I think that really starts to show the importance of these

different generations and also kind of hone in on when you’re trying to create something that’s clinically relevant

what you should be really focusing on yeah and I think you know that data suggest that like you said the first

generation clocks which are trained to predict chronological age they have less utility in predicting outcomes of Aging

then clocks like the second generation that are trained on other phenotypes of Aging or other biomarkers so this is

just another point to prove that we need to be talking about this especially in

epigenetic clock development for future development and refinement of these markers right this really guides kind of

where we want to go next in the field there has to be almost like some type of an agreement when we’re creating these clocks and again that’s why I like to

ask what are these best practices because then we may be able to come to an agreement and and you know follow

them or or train them in the same way so we can really start to get more of those Interventional uh trials and see what’s

actually moving the the clocks in terms of increasing and more relationship to disease or hopefully you know decreasing

so also BR can you just like in one sentence def Define like stochastic

aging versus Dynamic aging again just so I can make sure like we have that correct because I to be quite honest I

haven’t really read the paper fully but I’m still trying to wrap my head around

this idea of Aging yeah so stochastic aging is stochastic represents just random chance uh of of

accumulation of any kind of signal so stochastic aging is the marks that or

the Aging marks that you acrw randomly Dynamic aging or the Quasi

stochastic as we kind of call it is associated with some actual reason for

that aging to occur so this could be disease onset this could be you know exposure to environmental cues any kind

of external Factor that’s not random thank you that’s perfect that’s perfect and um they’re both changeable though

right yeah I mean the idea of stochastic uh aging is that it acrs just naturally

and so the I can’t say much of how much of stochastic change or stochastic aging

can be um changed definitely the dynamic can though right okay got it that makes

a little bit more sense again first generation clock measuring more chronological age it’s like natural aging there’s random chances that you’re

gaining you know methylation and and you know that’s affecting your overall aging this Dynamic aging there’s some actual

underlying reason whether it’s like a biomarker change or whether it’s like a disease type of change that it has in

terms of affecting the body yep yep awesome cool thanks for bringing that that paper up I know that that’s a

really exciting one and was just published today on the date that we’re recording it and I’m will definitely

link it out um all right so we talked about reproducibility including the

principal component analyses algorithms number two these epigenic clocks have to be related to clinical outcomes they

have to have an association to disease second third generation look better than first generation uh we now have even a

better argument behind that with this sto castic aging and this Dynamic aging

um if you had to pick one more because there’s a lot of other things you need to get to what’s another best bioin

practice that you would include I I don’t know if this is biotic but I would almost say this is more scientific and

kind of goes with the second one it’s validation I think the these clocks need to be vetted um not just based on you

know um white papers or like an N of one style um you know clinical cases but

really in these randomized control trials in these more large population

kind of meta anal not only meta analyses but just interventions because what I

think that affords is no one clock is built the same it’s inherent like when

you start to calculate a lot of these um outputs you’ll notice that the cpgs that

are shared is actually very minuscule really and so naturally even from a

numbers perspective the numbers are going to be different this is why the clocks and I I sorry can I go a little

bit off the rails here I understand that it’s really disheartening to see multiple numbers not agree with each

other but people need to understand these clocks are built for specific uses

and so if this clock is showing you as older or and this clock is showing you as younger there’s a reason for that

because it’s capturing a different part of aging that is saying you are older or a different part of aging that is

younger and so this is where and the reason why I’m going off the rails here is because this is where you need to go

back anyone you work with um in order to interpret these um these epigenetic

outputs you have to go back to the validated interventions you know this

clock is reduced in this context are you doing that that could be a easy way to

kind of start that conversation and so I think that validated interventions and I think the more um you know clinical

trials that these clocks are part of will then just open our eyes to like you know where does it work where doesn’t uh

where doesn’t it and I think that that’ll allow us to you know know more about where to utilize it

too yeah awesome awesome point and a good one to to end on in terms of I know there’s a ton of other things we can add

onto this list but number three validation and my followup is going to be well B what do you consider

validation but really you know these randomized control trials more clinical trials what can you do to like really

move these and change these did it work previously in animal models does it work to really affect our underlying biology

that one speaks to me just because I’m really helping people try and interpret these results on a daily basis and you

know I understand where the frustration and confusion may come from where and I know you said I know you said something

about like you know this being the last one I I and you touched on it there you also have to know which tissue type it

was built on because that’s you don’t want to use something that was built for Sal saliva on blood and vice versa so

just I wanted to make sure we added that too yeah four fourth one tissue type tissue type has to be consistent and we

we need to address that and we need we need to look at that um so no that’s that’s great that gives us a really good

Baseline Baseline on hey what are we actually looking at and you know it’s still way too early like we have a

couple randomized control trials and we’re getting really closer we have a lot of internal data true diagnostic that we don’t do a great job of sharing

but we will be getting better here soon um what’s you know really exciting though is yeah that we are still so

early like you know this this podcast it’s a lot of research it’s a lot of like far out there ideas almost and I do

imagine one day you know we look at this um 10 years from now and we just laugh

and say you know little did we know like we we thought we knew so much then there will be so much that we start to uncover

so I’m I’m just super excited about the the possibilities there I am too I I think it and I think that’s the reason

why you go into science or you do any of this kind of you know scientific work right um and I think that’s one thing

about about where we’re at at true diagnostic is that we understand in order to better provide for individuals

we have to understand the science first we can’t provide something without knowing how it was made or how it’s

operating and I think that with with that we feel the same way as actually a lot of the researchers do I I I look up

to a lot of like you know the Ricardo marionis of the world Steve Horvat Morgan Lavin uh Daniel bskis Terry

moffets abash Shalom kaspi like there are and if I didn’t mention any names I’m I apologize but um but like you know

we look up to these individuals because you like I think it’s a uh give and take

we are learning a lot of things we want to put out to the world but we’re also learning a lot from uh other individuals

that we’re trying to distill and and bring out to the people that are here at

true Diagnostic and also read our papers and things like that so I

I history has taught me that whatever I know now will be changed in the future of course so and I think at least

at least you know that like you already know more than a ton of people like if you’re open to that and you just accept that idea like you’re already like more

than halfway there I think that’s that’s science for you I would I would really want people to also feel that way too I

think the people that are listening to this podcast like I would hope that they kind of maintain that malleability the

neuroplasticity if you will to to kind of do that as well sure no good point

good point um okay this is I know I keep saying I feel like every time I bring up a new question or or topic like this is

the most important part or what I’m most excited ask about but this one really is probably the most exciting thing I want

to discuss today and it’s just the other use cases for epigenetics so I used to

describe like my company as oh we have you know the best biological age testing and that’s still how most people find us

and again where we can really make the biggest impact but that’s old news to me it’s it doesn’t get me exciting it

doesn’t get me out of bed every day um it kind of bores me a little because

there is this new idea of using epigenetics to report out other outcomes

and that is going to be what is called an EBP or epigenetic biomarker proxy so

it’s a new term it’s a completely new idea so run I you know you’ve been

leading this uh at true diagnostic can you just Define what is an EBP EBP was a

term that we um we kind of coined uh in the context of what we’re doing so the idea of an EBP is that we’re using DNA

methylation data to essentially impute or estimate values for proteins for

metabolites and for clinical biomarkers that you don’t actually need to take

that you don’t actually have to do traditionally and so using the same data set you can get estimates of those

values um this really came kind of on the heels of a lot of the work that Ricardo marioni has done in the past uh

where I think he coined his called episc scores and where he was able to um him

and his group were able to utilize DNA methylation to estimate

proteomics now how does this work I think it again it goes back to the central dogma we know that epigenetics

is connected to DNA DNA then goes to RNA RNA then goes to protein so they are relatively linked there is some shared

signal that is going from the beginning to the end and therefore and then also with that protein that elicits a uh

phenotypic or physiological response which is captured by those clinical tests it’s under that framework that that’s

why the logic of the ebps was kind of generated and so how we created this was

that we actually worked with um Harvard with Harvard University Brigham’s uh biobank specifically Dr Jess gaski Suz

group where we collected um samples from their biobank about 5,000 5,000 or so and then

calculated or Quantified their DNA methylation their metabolites of a subset the proteomics of a subset and

the clinical biomarkers of a subset and throughout a lot of machine uh machine learning modeling and a lot of that we

were able to create models where you identify which cpgs are most predictive

of that final outcome and so the reason why we did this initially was for aging

we wanted to build a multi-omic clock which took all of these um values and kind of Coes it or condensed it into a

single measure but what we started to realize is that even though we created that composite a lot of people are really

interested in how what is contributing to that composite score so how is this EBP actually affecting it because sure

you might be 76 if like chronologically but you’re 78 based on the on uh your

omic age so then naturally is like okay what what’s contributing to that then

you can start to go into the ebps because they are part of that calculation and then say Oh it’s that CRP that’s really you know not that

great you know it might show a percentile that’s like 90th percentile we can get to that a little

bit more details but that one is actually contributing is really high and therefore contributing to the overall

omic age your doctor your physician will Pro likely know or understand how that

CRP is being um is being affecting or how how to reduce that maybe and then

you start to actually gamify know exactly where to go uh in a more personalized uh approach

yeah that’s massive I mean everyone yeah should be on the the edge of their their seats right now right this is huge the

the implication of these epigenetic biomarker proxies and giving you guidance on what you need to work on to

lower your biological aging is is massive so right now there is reporting

that tells you what they are and what you need to work on to lower biological aging but I know furthermore we’re

actually developing reporting these out on a one to one ratio and

you know I I get my blood Strawn every six months like I’m sick of going to labc I’m sick of going to Quest I’m sick

of fasting and not eating until I get my bloods drawn and you know trying not to exercise in the morning and trying to

control for all these other external variables but what if you could just be

s a kit you’re sitting at home you take out the blood spot card you prick your finger with the lanet you fill up one

little circle that’s about the size of a quarter let it dry and you ship it to lab and you get results back I don’t

know in you know two to three weeks and it’s a very you know lengthy but

in-depth and important report that includes aging outcomes it includes clinical lab values it includes protein

levels and it includes metabolites like that’s what we’re getting at that’re we’re not too far away from that and

then it acts as a One-Stop shop with personalized micro does for recommended

supplements yeah so it really it really is insane so aside from even those those

EVPs verun I know you all are working on there’s like 4,000 of them that you can just like report out on from a couple

drops of blood is that correct sorry so um a apologies can you say that one more time because I think that this is a

really important question can you say that one more yeah yeah yeah so and again like this is me not being as

involved in you know the bioinformatics and the science that’s even happening at at true diagnostic so uh because we’re

always doing so many different things but essentially there are now these 4,000

EVPs I believe that we’re creating with Harvard to to report out on a one to one ratio is that correct so yeah so that’s

where we’re actually going towards and I will say this um and I want to make sure because I I I definitely am hearing what

people are saying sometimes your lab values and your EVPs may not 100% you know there might be differences right I

we have been working our hardest to get them to one to one but what we also start to realize is that these EVPs are

also picking up signals that are not even manifesting Downstream so meaning they’re not even coming out in the

clinical values like let’s say your CRP the reason the reason why I’m almost

pivoting a little bit is because of those 4,000 we’re actually running a study right now to understand why that

difference is there because previous work has shown that the CRP for example

at the end the clinical CRP is is not necessarily as Associated to age related

outcomes as the DNA methylation CRP is so that difference there is actually

attributed to more of capturing a different signal associated with aging compared to what the traditional lab

values are coming kind of coming out with and so right now at true diagnostic we’ve been able to expand our list from

396 that were originally described to about to close to 4,000 and it’s under

that pretense and why we’ve continued this way is because we’re starting to notice that there’s actually more

potentially more predictive power with these ebps than there were before so

even if there’s no there’s not an equal match even though we spent a lot of time to make them one to one still there’s

not always 100% match but that’s actually okay that’s good because now what you’re getting is a more predictive

measure for your age related outcome yeah exactly the CRP is a really good

examp a really good example because it was even done by an external group I I forget who exactly I link the paper but

where the Dame methylation CRP is actually more reproducible than the traditional CRP head-to-head and it’s

actually related to brain health outcomes where your traditional cpy is not so it’s like a 3 to six month

running average instead of just your plasma based biomarker at a specific point in time so a lot of people even

like to compare that DNA methylation CRP like the hba1c of inflammation so yeah

that’s a really good example and a super exciting one I already know like that part of the the podcast and conversation

will be my favorite of this entire episode um because there again is so many things we’re finding out but so

much we don’t know and then we’ll actually be able to go back and look at each individual EVP and say hey what is

this really telling us what is this really capturing I mean it’s not necessarily it’s novel it’s capturing

something that we probably didn’t even we should be looking at as it relates to our health but it’s probably Associated

to a lot of other outcomes downam and and if and this is the thing like obviously you can do all of this in one

go that’s fantastic but if you are really inclined to understand what your clinical outcome is we’re not stopping

you from that we’re just saying that what we’re what we’re essentially measuring is likely showing you a signal

that is first clinically relevant but two maybe more predictive and then if you really want to figure that out later

on you can then also still do the clinical out outcomes that’s that’s on that’s privy to

you yeah not not not an end all be all like replacement test that’s not what we’re saying we’re saying do them all

together um and um you know we’ll we’ll be able to tell you a little bit more about what those actually squeeing a little bit more juice with us I’m just

saying yeah yeah exactly now um just really quickly we’re we’re kind of

getting to to the end here I know on a previous episode I talked about methylation risk scores or mrs’s this

was Dr Michael Thompson out of UCLA um and you all can listen to that episode

if you really want to take a deep dive into his particular study but ver just

so if if people are you know listening to these in order like what’s a methylation risk organ and then can you

just briefly explain like how does that compare to an epigenetic biomarker or proxy or EDP yeah so essentially the way

that we trained or generated these epigenic biomarker proxies were identifying cpg’s markers that were

highly Associated to let’s say a protein value you can do the same thing directly

to just disease onset so let’s say this individual got has cardiovascular

disease this person doesn’t there is a statistical method a machine learning method where you can identify which cpgs

can actually quantify or predict sorry the onset of that um of that disease and

so this is kind of using um Cox proportional Hazard models for those scient sorry statistically inclined um

and and ultimately it’s rather than using age as a surrogate to estimate the

risk of that disease happening which is typically what we kind of do with OMC Mage is that we calculate OMC Mage and

then you know compute a disease that comes out of it like a um estimated

disease outcome why not just go directly to the disease itself and so it doesn’t negate anything about

OMC MH but what it does is that allows a a direct comparison to the disease outcome and so that is what essentially

that methylation risk score is it’s a risk of you know that of that that disease or whatever that

phenotype that you’re looking at or disease comorbidity whatsoever and predicting that based on directly from

the methylation marks that are Quantified by DNA methylation yeah so it’s act your actual

risk and not to confuse listeners but I think you know we were even going to call our EVPs originally mrss scores

methylation risk scores but it just didn’t feel right it’s not a risk it’s not a disease risk so it’s an epigenetic

biomarker proxy I love it it’s an EVP it’s it’s using epigenetics d methylation as the input biomarker and

acting as a proxy for some other type of output and I remember I was in uh I was

at ardd when we changed the name too so everyone started saying EBP E I was like

what the heck is the EBP but then I started to realize yeah you’re absolutely right we can’t call these methylation risk scores like these are

proxies of that actual marker and so I’m actually very I’m I was ecstatic when we

changed it to ebps truth be told yeah yeah cool to coin coin that term um cool

Veron um getting to the end here what are other areas of discovery that excite

you I know there’s like so many things we’re we’re working on um is there a specific project like I don’t even know

this personally about you that you’re excited about at the moment while you get um I mean I’m living the dream right

now so it’s kind of it’s kind of hard um I mean I was fortunate enough to be on

this project looking at Twins and the effect of diet on twins um but there’s

you know I think I was speaking to Michael Corley from Cornell and we were

talking a little bit about this project and one thing he raised up was we actually don’t really know a lot about

methylation patterns among twins um I mean there have been studies but like you know not a lot has been done and so

I think kind of going into the whole um the the twin aspect has been interesting just because I can’t think of any more I

can’t think of a study design better to test how environment impacts um

individuals you know from from a context of a genetic and just also environmental

um uh cues obviously doing that kind of an experiment is very tough um they need

to adhere to a lot of uh compliance uh as shown in this paper but um that’s

like one thing that would be really awesome number two is I’d love to a lot of where we’re moving towards is kind of

satiating in my head this aspect of connecting molecular to Output the

physiological output how much of what we’re seeing on the molecular side actually translates to clinical outcome

and I think that you know really understanding that we’re starting

to really scratch on that surface right now but I don’t but I I I think there

could be so much more done like for example whoop like even from a technical side if we can connect our methylation

patterns to let’s say whoop outputs what’s conserved because I think some people would say there’s shouldn’t be

anything different but there might be and I think that that’s that’s really the area that I would love to kind of

see you know foray into and like we might yeah yeah I think we’ll get some of that wearable data integrated

for for sure later on we’re we’re definitely working on on some things there so I I would be excited to see those those correlations or those

connections too yeah and it’s yeah yeah so like I said it’s it’s a lot of stuff that we’re we’re working on already but

that’s I think it’s Guided by this curiosity this satiation of the question

um I’ve been always very question oriented and so I think that um yeah so

these questions the more papers and the more studies that we do there’s a lot

more questions that come out of it and so um it may be never ending but I think we’re getting there we’re getting to

some kind of end I think that’s yeah that’s the important part that we’re we’re having

more more questions then then answers more more R&D more more Discovery so last question Brun this is a curveball

question if you could be any animal in the world what would you be and why oh my God I mean I’m going to stay away

from like the usual answer which is a dog cuz I love dogs I have Robbie is right back there somewhere my dog um

H okay a crow and hear me out a crow a

crow crows are intelligent they’re very intelligent they have the ability to fly

which I I don’t know flight is one of those things that’s like anytime I’m on a plane I’m Giddy and I’ve been on

planes for the last like 25 years um but I think it’s their intelligence they’re ability to utilize

their environment for their own you know their own uh need is incredible and as

much as there you’re right they’re really smart as much as they’re Associated to like you know negative context in a lot of you know societies

and religions and all that I think from just a biological perspective they’re so intelligent and they’re so crafty that I

mean I they imitate they imitate people too exactly yeah i’ take Pro yeah

they’re really really smart that’s interesting most people yeah have like anxiety around flights and and things

like that too I never knew that about you so I’ll I’ll be paying extra extra attention when we we fly together next

time I mean I still hate turbulence I absolutely hate turbulence oh yeah no one likes turbul

turbulence uh cool we’ve come to this uh amazing podcast um interview in terms of

the the ending for listeners who want to connect where can they connect with you where can they find you and how can

researchers access these these absolutely yeah so I think I’ll start with the organization so true diagnostic

uh www.tr diagnostic.com um and also on the Instagram uh true diagnostic

official for me personally U you can find me on LinkedIn Veron darara V Ru n

DW a r AA I’m on in uh LinkedIn and Instagram as well same just it’s my name

um and then in order to access a lot of these biomarkers I highly implore everyone to kind of go through the

papers that we published and then also contact me directly Veron true diagnostic.com because we are more than

happy to work with other researchers and other Physicians on you know implementing a lot of these things I you

know definitely go Hannah has been killing it killing it in terms of you know for forefronthealth.com

like you know how any of these things so email Instagram LinkedIn yeah catch mewhere and all them all the cool we’ll

link them out well thanks everyone again for joining us at the everything epigenetics podcast and remember you

have control over your epigenetics so tune in next time to learn more thanks