Hannah Went:
All right, welcome to the Everything Epigenetics podcast. Zach, thanks for being here with me today. I’m excited to chat with you.
Zach Kaminsky:
Yeah, it’s a pleasure to be here.
Hannah Went:
So I gave our listeners, you know, a brief introduction. I pre-recorded, you know, put that in the beginning of our episode, but I want to hear it from you. You know, I’d love to learn a little bit more about your journey, where you ended up, how you got to where you are today. So can you just talk a little bit about that, how you got involved in this space and the studies you’re doing now?
Zach Kaminsky:
Sure, yeah, absolutely. It’s a bit of a circuitous route that brought me to epigenetics, but effectively, I think I have to rewind back to the beginning, which is that I met a girl.
Hannah Went:
Huh?
Zach Kaminsky:
I’m in Japan teaching English, who was Canadian, and we fell in love, and then moving back to, I’m from Baltimore, Maryland originally. I learned that you can’t just move to Canada from the US if you want to. So I had to get a visa, an after visa, that would get me up there. So the only thing I could do, I had some lab science experience and I was going to take any job that came to me and through some connections this person starting an epigenetics lab reached
Hannah Went:
Thanks for watching!
Zach Kaminsky:
out. At that time I was a technician at Hopkins and I recall some of the doctors that I was working for obviously with a bit of a conflict if they wanted to keep me. They were like, oh, epigenetics in psychiatry, are you sure? Really don’t think that’s going anywhere and I was like,
Hannah Went:
Hmm.
Zach Kaminsky:
I don’t care. I just need to get up to Toronto so, um, I took this job and Really thrilled that I did we I was under art patronus who is really a visionary in psychiatric epigenetics He was doing a lot of the the groundwork to really push the the theories that epigenetics was important in psychiatry And so I was his Technician for a while and then eventually, you know entered a graduate stream that girl who is now my wife, Sharimar Kaminsky, basically said, you know, you should do graduate school. And so I went through the paces there and found myself graduating, having done a PhD in a field that was just becoming hot, a field that I had a lot of training in. We helped to create some of the first microarray technologies, challenging and technically noisy as they were, we were really sort of, I think on the forefront of trying to investigate the importance of epigenetic in a number of biological situations, including psychiatry, but other things like twins and things where environments and stochasticity of methylation changes over time were being figured out. So that’s really
Hannah Went:
Yeah.
Zach Kaminsky:
the long story of how I got into epigenetics. I stumbled into it and my wife, to her credit, suggested
Zach Kaminsky:
I try to become more professional.
Hannah Went:
Ha
Zach Kaminsky:
And
Hannah Went:
ha.
Zach Kaminsky:
I’m glad I listened to her.
Hannah Went:
Yeah, well, I love those stories that are always, you know, they kind of happen by chance and they’re random and I’m a sucker for a good love story. Those are like I love the cheesy, you know, rom com movies and whatnot. So, no, I really like that. And that has to be really interesting and fun and more of like a discovery phase that you’re you are on the forefront of this new field and just kind of taking a leap of faith and really, you know, being a mover at the beginning of it. So that’s what I’m excited to, you know, obviously talk about and chat with you today. Your current work focuses more on epigenetics and psychiatry, but more in particular you really specialize in suicide, PTSD, and postpartum depression, which again is going to be our main focus today. But what do you, can you talk about some of the work you’re doing now and to a further extent? And maybe tell us about your favorite too. I’m always curious to hear about what your favorite is to study.
Zach Kaminsky:
Yeah, so, gosh, what are we doing now?
Hannah Went:
A lot.
Zach Kaminsky:
Lots of things, you know. We’re still focusing on suicide epigenetics, but really focusing now on postpartum depression epigenetics
Hannah Went:
Mm-hmm.
Zach Kaminsky:
is what I’m excited about now. With epigenetic changes being so variable and being influenced by changing environments, The really trouble with studying epigenetics is that a lot of the times the disease has already occurred. People are already under treatment, right? But postpartum
Hannah Went:
Mm-hmm.
Zach Kaminsky:
depression is one of those few psychiatric diseases that you know when it’s likely to happen, which is, you know, after giving birth to a baby. And so, you know, that gives you an opportunity to measure
Hannah Went:
Mm-hmm.
Zach Kaminsky:
the blood or whatever you want to measure with epigenetics, whatever tissue, And so, you know, I’m really excited about working on that because about maybe 11 years ago or 12 We’ve discovered a set of epigenetic biomarkers that were perspectively predictive of postpartum depression right now. I have a I’m a co-founder of a startup company Dionysus digital health. That’s trying to really market these Biomarkers bring them to the forefront make them them useful So that’s exciting because I remember a decade ago or whenever when we first published this, as a younger scientist then, and those were my first news appearances, postpartum depression biomarkers discovered. And I was at Johns Hopkins as well, my laboratory there. And when you’re at a place like that with a lot of clout, when you announce a discovery, you get a lot of press.
Hannah Went:
Oh yeah.
Zach Kaminsky:
So here I was thrust on the world stage making news articles come out and basically they’re like, well, when is the test going to be?
Hannah Went:
and…
Zach Kaminsky:
And I have no idea. So, but we’ve been studying it now, continue to expand our understanding of this for like the next 10 years. We’ve done a lot of replication studies. We’ve done neuroimaging with these biomarkers to try to get into the brain, so to speak. We even have some interesting human models estrogen has been knocked
Hannah Went:
Thanks for watching!
Zach Kaminsky:
down with a pharmacological agent and then you know our biomarkers predict who’s gonna get depressed After that
Hannah Went:
Wow.
Zach Kaminsky:
happens, you know, it’s really exciting. That’s unpublished data, but you know, we’ll get it out there eventually So yeah, now it’s really exciting to finally be trying to move that forward
Hannah Went:
Yeah.
Zach Kaminsky:
into being out there to people
Hannah Went:
Definitely. There are all these, I get all the Google alerts, I’m set up for all the alerts from all the journals and whatnot, and every time a paper comes out, I get so excited, especially as it relates to pregnancy and these different markers, and I always think in my head, too, oh, this is such a cool finding, but when will it become commercially available, and we can start testing for this. I think people are really excited to start getting this in their own hands and being able to identify markers and find maybe different interventional therapies that work for them. So I found it interesting, too. rather obvious, but like you said, the timing for the postpartum depression, that’s nice. Epigenetics is a good way to measure it because obviously, yeah, we know when that’s going to happen. But in my studies and kind of what I spend my time looking into is a lot of the interventional trials with epigenetics and it’s really hard to say, okay, when do you get the peak effect from the intervention? And then number two, how long does that intervention last? So it’s really hard to time. You know, obviously you get a baseline, but the after kind of epigenetic tests. So I think we just need more and more of that. know until we start to measure so that that can be a little frustrating in the field.
Zach Kaminsky:
Yeah, and it’s not cheap to run, you know, if you want to serially
Hannah Went:
Right.
Zach Kaminsky:
test things. I have a collaborator who does sleep research and, you know, she was
Hannah Went:
Mm.
Zach Kaminsky:
saying, oh, we could do like circadian rhythm. And, you know, at first you’re like, yeah, yeah, we could do that. Like I have samples from, you know, multiple time points, you know, over,
Hannah Went:
Mm-hmm.
Zach Kaminsky:
of the day, over multiple weeks. And like, I’m just thinking, oh gosh, this is gonna be really pricey. Even if we use, you
Hannah Went:
Yeah.
Zach Kaminsky:
know, pyro sequencing, the least expensive sort of methylation assessments. But costs can be prohibitive when it comes to these sorts of technologies, especially with genome-wide technologies that are a bit more agnostic but give you a better chance
Hannah Went:
Right.
Zach Kaminsky:
of finding whatever is real.
Hannah Went:
Yeah, definitely. Yeah. And the interventions I was kind of mostly hinting at or thinking about in the back of my mind are like the stem cells and the exosomes, because then it’s like all about timing and kind of the type and differences. So yeah, it can become very, very pricey and hard. Well, um, you know, you’re, you’re a specialist in this space. So, and I want to hear it from, from an expert and dig a little bit further into just why is it important to study these subjects, you know, the postpartum depression, uh, the PTSD, the, the suicide, you know, um, what, by, looking at epigenetics, what can we learn about those through that lens? I know that’s rather obvious, again, but would rather hear it from you as an expert.
Zach Kaminsky:
No, no, it’s a great question because I think we need to back up and think, all right, yes, why epigenetics? And the big why, which I think has remained constant for a little while, is because there’s hope in epigenetics that we’ll find something that we haven’t found elsewhere. So by contrast, pure genetics, looking at DNA sequence, that had a lot of hope behind it when the genome was sequenced. NIH spent millions upon millions of dollars trying to do genome-wide association studies and say, we’re finally going to understand and solve these diseases like schizophrenia, depression. And, you know, it hasn’t really panned out. Those are, you know, they’re complex non-Mendelian diseases. And as my supervisor back in the day, Art Patronus, the visionary, would say, you know, these complex non-Mendelian features, having twin discordance, for example. Just having age of onset be right after puberty or onset like postpartum depression after hormonal fluctuations that happen during pregnancy. Typical genetics can’t explain that, but the epigenome really sits at the intersection of genes and environment. It can be modified by the environment. It can be changed by the genome. of nature and nurture, right? So by studying epigenetics, can we find things that we haven’t been able to find before? And so, you know,
Hannah Went:
Mm-hmm.
Zach Kaminsky:
I set my program about doing this in, you know, when I started my laboratory at Johns Hopkins, in also looking really for biomarkers. There’s sort of two sides to using epigenetics. One is understanding the etiology, which can be
Hannah Went:
Mm-hmm.
Zach Kaminsky:
confounded by the sort of cause effect nature the fact that these things can change, right? So if you’re studying people who are depressed, they’re likely on antidepressants. We know that these can change, you know, histone deacetylase, in histone deacetylases
Hannah Went:
you
Zach Kaminsky:
and methyltransferases that in turn can affect other epigenetic marks like DNA methylation. So is what we’re finding, you know, causative of the disease or
Hannah Went:
Thank
Zach Kaminsky:
is
Hannah Went:
you.
Zach Kaminsky:
it
Hannah Went:
Thank
Zach Kaminsky:
a,
Hannah Went:
you.
Zach Kaminsky:
you know, consequence of treatment or of the disease itself? So, you know, epigenetics in that way, but it also can be modified by a lot of these epidemiological factors. These factors, like being small for gestational age is associated with a number of mental illnesses, or having early life trauma predisposes to depression and suicide, or PTSD may be a little bit different depending on the timing of that trauma and in complicated ways. samples and thinking carefully about all of these potential confounders and where does my data, you know, what is it going to be telling me, what is it useful for? It allows you to really ask the question, you know, is epigenetics important? And so far, you know, we think it is. So we’ve found a number
Hannah Went:
Mm-hmm.
Zach Kaminsky:
of really, you know, replicating interesting findings. I should mention replication, you know, is another important thing that didn’t necessarily happen a lot in genetics, but we had a finding in suicide at the SKA2 locusts at SCOT2
Hannah Went:
Mm-hmm.
Zach Kaminsky:
that has since been replicated in a number of different papers. I think there’s at least been now something like 10 or 11 independent
Hannah Went:
Oh wow.
Zach Kaminsky:
papers. Some of them may have been from us, so maybe it’s only like seven
Hannah Went:
Ha
Zach Kaminsky:
other papers,
Hannah Went:
ha
Zach Kaminsky:
but a lot. external replication
Hannah Went:
Mm-hmm.
Zach Kaminsky:
because lets you know okay, you know there is something to this. This is this is really not likely to be chance, so
Hannah Went:
Yeah, yeah, that validation work. No, I think that was beautifully put and explained. So thanks for that explanation. I really do think epigenetics is the new frontier, the new genetics, this revolution that we’re seeing. And yeah, the challenges along the way, I think, is what makes it really fun. We’re gonna run into that in any type of field. I think what, there’s 28 million different methylation markers in every single cell type.
Zach Kaminsky:
Thanks for watching!
Hannah Went:
So we don’t know what we don’t know. We’re still learning. And I think that’s what makes the possibilities really, really unique with this field as you were hinting with the genetics is more stagnant and didn’t have a lot of replication studies there. So you led me right into my next question though. I was going to ask you about that study, wanting to focus on your work more in the stress and the suicide realm. So you really focus more on these mechanism of actions too, which I think is really unique about your work. But that paper, I’m going to name the title is Stress Vulnerability and Epigenetic Variation of SKA2. So what did you do there? What does that mean? And maybe describing that SKA2 to the listeners as well.
Zach Kaminsky:
Yeah, sure. Yeah, so, you know, SKA2, SCA2, for easier pronunciation, you know, we really found this doing an agnostic EWAS, a genome-wide association, epigenome-wide association study, and SCA2 came to the top. You know, we started with a potentially disease-relevant tissue for suicide. We started with the brain, brain tissue, and actually fax-sorted nuclei versus non-aromal or glial nuclei. And SCOTU, you know, was one of our loci that came to the top. So, you know, we were excited about this. Actually, you know, there were four loci that came to the top. But SCOTU was the only one that made sense biologically for potentially being involved in in suicide because it’s one of these sort of at first look not necessarily very interesting microtubule genes. But we noticed that a little bit of work had been done showing that with a yeast-2 hybrid assay, it associated with the glucocorticoid receptor, which is a big player in depression and suicide because it’s the brake pad for the HPA axis. This is our normal stress response. So we know that stress is important for everything from crossing the road to going to a meeting, But, you know, when the stress is over, glucocorticoids bind to the GR, glucocorticoid receptor, and they help this then transactivates into the nucleus, and it expresses various other genes that help shut off stress, right? So it’s the brake pad of the stress system. SCA-2, that microtubule, that scaffolding protein, it’s a chaperone. It’s basically like the train car that brings GR into the nucleus to do its job. And when you knock it down with siRNA, GR can’t get into the nucleus. It can’t transactivate.
Hannah Went:
Mm-hmm.
Zach Kaminsky:
And so we think it’s a molecular mechanism for what we see in folks with depression and suicide, that they can have a stress response that is extreme
Hannah Went:
Mm.
Zach Kaminsky:
and then doesn’t shut down. And it really fits into leading biological theories
Hannah Went:
Yeah.
Zach Kaminsky:
for suicide, which is diathesis stress. It’s an underlying biological vulnerability,
Hannah Went:
Thanks for watching!
Zach Kaminsky:
like a brake pad on a car that is not driving, isn’t a problem, right?
Hannah Went:
Mm-hmm.
Zach Kaminsky:
But add that stress.
Hannah Went:
Yeah.
Zach Kaminsky:
Now your car is driving, and you’re unable to shut off the stress when you need to. You’re unable to break that car, and you have this sort of positive feedback mechanism, which results ultimately in rewiring of the brain.
Hannah Went:
and I’ll see you next time. Bye.
Zach Kaminsky:
So areas in the frontal cortex involved and decision-making
Hannah Went:
Mm-hmm.
Zach Kaminsky:
are less active, and there’s lower functional connectivity to sort of top-down control on the amygdala, our fear
Hannah Went:
Mm-hmm.
Zach Kaminsky:
and anxiety center of the brain, right? So now we’re going from a sort of epigenetic finding to an intuitive understanding of like, hey, if I’m not able to shut down impulsiveness, and I have like elevated fear and anxiety, you could start to
Hannah Went:
Mm-hmm.
Zach Kaminsky:
envision how this might lead to suicidal thought.
Hannah Went:
Yeah.
Zach Kaminsky:
And so, you know, we were able to look in some of our functional connectivity studies that the epigenome signatures it’s got to were effectively associated with fMRI signals there. And other groups in some of those multiple papers I mentioned have shown similar
Hannah Went:
Mm-hmm.
Zach Kaminsky:
things. They’ve shown structural changes in, I think, the frontal pole and functional connectivity changes as well. of epigenetic factors being important for potentially leading to the pathophysiology of psychiatric disease. Now, one question, you know, there’s still lots of things we don’t know, like how does
Hannah Went:
Yeah.
Zach Kaminsky:
it get there?
Hannah Went:
Yeah.
Zach Kaminsky:
And I can share some unpublished speculation based on unpublished data. So I mentioned to you before the podcast that we have an ice storm today. Actually, Canada gets these sometimes.
Hannah Went:
Mm-hmm.
Zach Kaminsky:
I’m up in Ottawa. disaster was the Quebec ice storm. I think it happened in 1996 or something. And there’s
Hannah Went:
Mm-hmm.
Zach Kaminsky:
been a methylation study done on the offspring that were about 13 years old. So what you know we were able to because there’s a genome-wide study we’re able to download the data and look at the maternal distress scores and we find the sort of risk epigenetic signature in Scott who associates with how distressed a mother was during pregnancy changes their scotum methylation towards the risk phenotype. So this could be one of those
Hannah Went:
Wow.
Zach Kaminsky:
sort of epigenetically passed on, I don’t want to say inherited because we can confuse what goes through the germline versus,
Hannah Went:
Mm-hmm.
Zach Kaminsky:
in this case it would be cross placental epigenetic
Hannah Went:
Like that,
Zach Kaminsky:
transmission.
Hannah Went:
yeah, transgenerational epigenetic inheritance, maybe.
Zach Kaminsky:
Right, yeah, yeah,
Hannah Went:
Right,
Zach Kaminsky:
so
Hannah Went:
bird? Okay.
Zach Kaminsky:
I had the good fortune, I was sitting with Emma Whitelaw, who’s a famous epigeneticist, and she was saying, we were just waiting in a hotel lobby, and she was like, I want to
Hannah Went:
Yeah.
Zach Kaminsky:
coin the term transgenerational epigenetic inheritance, which goes through the germline, and transgenerational epigenetic effects, which don’t.
Hannah Went:
Oh,
Zach Kaminsky:
So I think this, so
Hannah Went:
yeah.
Zach Kaminsky:
anything, like early life trauma is a transgenerational epigenetic effect, because it
Hannah Went:
Mm-hmm.
Zach Kaminsky:
goes through, know, a mom could be depressed because of her epigenome and then pass that to her kids by being neglectful during a developmental period or
Hannah Went:
Right.
Zach Kaminsky:
generate enough cortisol during pregnancy it crosses the placenta that that epigenetically reprograms the child. So those are the transgenerational epigenetic effects according to Emma Whitelaw. I don’t
Hannah Went:
Hehehe
Zach Kaminsky:
know if it took off or if anybody else calls it that but yeah I don’t know.
Hannah Went:
Yeah, I know Dr. Michael Skinner out of Washington State University, he does it in mostly animal-based models, the transgenerational epigenetic inheritance, but I like the effect, the one where it’s, I think it’s called something else too, where it’s not passed through the germline, but I like that. I’m going to have to look more into it in her work. But no, that’s great. I mean, that’s so interesting to talk about and to think about kind of, again, the mechanism of action and how that plays through. So I’m going to ask a question. We don’t know, but can you do anything about it yet? Do we know any hints, any information?
Zach Kaminsky:
Oh, can we modify or change the epigenetic code at SCA2, for example?
Hannah Went:
Y-yes, yes.
Zach Kaminsky:
Yeah, we don’t know, you’re right, we don’t know.
Hannah Went:
Yeah.
Zach Kaminsky:
Yeah, I mean, it would be really interesting to sort of effectively look at that. It’s one of the challenges of studying epigenetics is these things can be malleable, but we don’t know how
Hannah Went:
Mm-hmm.
Zach Kaminsky:
malleable genetic changes are. I mean, we know that a little bit of work has been done. Let’s see, Chloe Wong, who was at John Mills group and then started her own group, did some of the studies showing that cognitive behavioral therapy was able to, I think, reverse FKBP5 DNA methylation post-therapy.
Hannah Went:
Mm-hmm.
Zach Kaminsky:
So, you know, she may have done another one in a serotonin
Hannah Went:
Yep.
Zach Kaminsky:
these things can be modified, which would make sense, but we haven’t looked, so.
Hannah Went:
Yeah, yeah. Well, no, that’s a great paper. I love it. I’ll link it in case anyone who’s listening wants to read a little bit more. I highly, highly encourage. One more question kind of about maybe your work in that realm. What about cell type heterogeneity in the brain? I know we were kind of chatting over that via email. What, can you talk about what role that plays or may have played in that paper?
Zach Kaminsky:
Sure, yeah, yeah, absolutely. So, you know, this is one of the interesting things about epigenetics is that it helps to define the over 200 cell types in the human body. But that it also, I was mentioning earlier that it can be challenging to study. And that’s one of the reasons why is because anything that has a change or a different proportion of cell types, so we could imagine something like neurodegenerative diseases like Alzheimer’s disease, if you’re looking at the brain, could have a different proportion of neurons to glia. Or if you had neuroinflammation, as you may expect to see in depression or suicide, then you’re going to get more glial cell types. And what we found, part of my first project at Johns Hopkins, my first R21 grant, was to really fax sort or isolate the neuronal and glial nuclei, which isn’t super easy, but I learned the technique
Hannah Went:
Yeah.
Zach Kaminsky:
from Sharam Akbarian. and was able to pull it off at Hopkins and do epigenomic profiling there.
Hannah Went:
Thanks for watching!
Zach Kaminsky:
So one of the things I’m digressing a little bit because
Hannah Went:
No.
Zach Kaminsky:
we did that in the SCOTU story as well to find SCOTU by reducing the heterogeneity. But we were also able
Hannah Went:
Thanks for watching!
Zach Kaminsky:
to create a bioinformatic tool that tries to control for this in other people’s brain tissue samples. So you can already
Hannah Went:
Thank you.
Zach Kaminsky:
have
Hannah Went:
Thank
Zach Kaminsky:
done
Hannah Went:
you.
Zach Kaminsky:
your brain-related study and found various things,
Hannah Went:
you
Zach Kaminsky:
but the idea is you then apply this model and you get a proportion of neurons to non-neurons that you can use as a covariate to sort of control for and understand if those hits are being driven just by cell heterogeneity. So this is really important in the brain when we’re considering etiology. I’m kind of on the fence whether to apply this. So Dr. Hausman did this in blood. white blood cell types, natural killer cells, B cells, CB4 and A, T cells, granulocytes, and came up with an elegant model that gives you proportions of these as well that you can use as covariates. But I mentioned earlier, you know, there’s sort of, we can look at etiology for psychiatric diseases or we can try to leverage biomarkers. And I’m always on the fence about whether to apply these covariates first when I’m looking at blood. And I usually don’t. And the reason is because So I just want a biomarker. If I want something that is indicative of that disease state, in a lot of psychiatric diseases that’s inflammation. So we may expect to see differing cell profiles due to this sort of underlying disease pathophysiology.
Hannah Went:
Mm-hmm.
Zach Kaminsky:
That’s why we would see it in the periphery. And so I don’t want to necessarily normalize that away. I need to be cognizant of that and check, hey, you know, I found these interesting genes.
Hannah Went:
Mm-hmm.
Zach Kaminsky:
Are they just because of cell type heterogeneity? the end of the day, if that gets you a blood test
Hannah Went:
you
Zach Kaminsky:
that allows you to test for a disease, you know, then that’s fine. You’re sort of in good shape. So again, being cognizant
Hannah Went:
Sure.
Zach Kaminsky:
of cell heterogeneity is always important, but you don’t necessarily always want to just normalize it away without understanding if there’s an effect that you can glean as a proxy for something going on in your disease of interest if you’re doing it by yourself.
Hannah Went:
Yeah, I think that makes sense if it applies to, you know, what you’re looking at or if it doesn’t in your case, I understand kind of the backing and reasoning behind, you know, choosing just the one bio marker for the psychiatric disease. So no, that makes sense. Appreciate that answer there. One thing, and I found a lot of news articles when I was writing up this agenda that I found about your work is, and we were kind of talking about, you know, pushing this to application, pushing this to commercialization. So something I found fascinating is you actually created a machine learning approach that predicts future risk to suicidal ideation from social media data. So tell us more. What did you do there? What did you create? This is like an awesome product. Again, I was reading in a bunch of different articles.
Zach Kaminsky:
Sure, yeah, no, happy to chat about that. Yeah, so, you know, I was at Johns Hopkins, but my wife, I mentioned at the beginning of this podcast, is Canadian and, you know, kind of always wanted to move back to Canada. You know, being from Baltimore, we moved there for a number of years. We thought we’d get more babysitting than we did for my kids from my parents while there. But eventually, I got recruited back to Canada as the Chair of Suicide Prevention Research. I should say the DIFD Matt Ganslin Chair of Suicide Prevention Research, and in down share they fund me and my position here. So, you know, this was really with a mandate to sort of address suicide in the area. And
Hannah Went:
Thanks for watching!
Zach Kaminsky:
I was thinking of that stress diathesis model that I mentioned earlier, where you have an underlying biological vulnerability, like a brake pad for the stress system might not be working, but it has to meet a time of stress. That car has to be driving. And so how do we measure that stress in real time? I didn’t think
Hannah Went:
Thanks for watching!
Zach Kaminsky:
that epigenetics alone was going to be enough. So I set about
Hannah Went:
Mm-hmm.
Zach Kaminsky:
delving into a bit of natural language processing work. I love to code. And
Hannah Went:
Oh.
Zach Kaminsky:
when you’re between two institutions, there can be a
Hannah Went:
Hehehe
Zach Kaminsky:
great period where you haven’t started one job and the other one is kind of winding down. So I use that as a bit of a sabbatical to learn natural language processing techniques in Python. And so what this does is it takes social media posts, anything text-based, and it converts like hopelessness.
Hannah Went:
Mmm.
Zach Kaminsky:
So, we use Twitter data because it’s convenient and public. How hopeless is this tweet? What is the burden score for this tweet? Is this a depression related tweet? What number between
Hannah Went:
Mm-hmm.
Zach Kaminsky:
zero and one is the sleeplessness of this tweet? And so, we end up with a matrix of data for each tweet and then we can apply machine learning methods to predict suicide based in people on Twitter versus not. And so what we found was that this was really predictive of people’s impending suicidal thought into the future. So
Hannah Went:
Mm-hmm.
Zach Kaminsky:
the machine learning was able to pick up a pattern that preceded people expressing suicidal thought on Twitter. And we’re able to look at suicide attempts as well, which is a more severe
Hannah Went:
Mm.
Zach Kaminsky:
form of suicidal behavior. So we know that more people think about suicide than act on it, and more
Hannah Went:
Mm-hmm.
Zach Kaminsky:
people act on it than die by suicide. So, you know, being able to predict suicide attempts as well was really quite interesting. Because Twitter data, you know, you can sort of go back in time, you can sort of look at these profiles ahead of time. So you can look for,
Hannah Went:
Mm-hmm.
Zach Kaminsky:
oh, I’m coming out of a suicide attempt, I was at the hospital, and you can look at the patterns before that. So yeah, we were excited that our I think 85% accuracy
Hannah Went:
and well.
Zach Kaminsky:
area under the curve, if I recall correctly. So that’s about, I think 85% sensitive and something like 80% specific
Hannah Went:
Yeah.
Zach Kaminsky:
for suicidal thought. Yeah, so that was an interesting technology. And because it’s digital, you can do it pretty cheaply and quickly.
Hannah Went:
Yeah, very cool. Is this still being like used today and anything or are people using that then?
Zach Kaminsky:
Yeah, so, you know, in my role here, we’re trying to think of neat ways to use this.
Hannah Went:
Mm-hmm.
Zach Kaminsky:
So I’ve come up with a couple ideas for a way that you could have a suicide intervention app where you could
Hannah Went:
Mm-hmm.
Zach Kaminsky:
have these scores and you could share them with trusted confidants, and then if your score
Hannah Went:
Mm-hmm.
Zach Kaminsky:
was ticking up, you know, those confidants could reach out to you. One of the things with suicide is it’s a low base rate event. Even suicidal thought
Hannah Went:
and I’ll see you next time.
Zach Kaminsky:
is a
Hannah Went:
Bye.
Zach Kaminsky:
low base rate event, which means that no matter what, false positives, even like a perfect test, 99% sensitive and 99% specific with 11 suicides per 100,000, that means a thousand false positives. So we’re not like sending anyone to the hospital
Hannah Went:
Gotcha.
Zach Kaminsky:
based on this, but if your friend saw this data and they reached out and they asked you to play basketball, you know, that would probably be a pretty tolerable intervention, right? And we know that connections
Hannah Went:
Yeah.
Zach Kaminsky:
are protective. So you know, can we leverage this in a way that’s smart for the field and does having allow you to act at a time when you don’t necessarily need crisis response. You know, you just need a few more of those protective factors. That’s the theory at least and we’d like to be able to use it that way.
Hannah Went:
Yeah, yeah, well, I love it. I think that’s great. I think we need more assets like that, somewhere to start and build off of it. So I like the apps and things that you’re thinking of creating and moving forward with. So I’m gonna have to follow along there and see what you come up with. I’m gonna switch gears completely. I appreciate your chat with me about the suicide and the PTSD-based model and how that’s already pushed to application with your Twitter application there. So I wanna move the conversation more I know we chatted a little bit about it in the beginning.
Zach Kaminsky:
Sure.
Hannah Went:
You have a lot of work there. You have several papers that investigate how DNA methylation biomarkers can predict both antenatal and postpartum depression. So can you describe just a little bit more in detail some of those studies and what you found there?
Zach Kaminsky:
Yeah, absolutely. So, you know, we’ve talked a bit about the challenges of studying epigenetics in peripheral tissues.
Hannah Went:
Yep.
Zach Kaminsky:
And so, you know, to try to get around this, the way we started this study was to start with a mouse model. We actually
Hannah Went:
Mm-hmm.
Zach Kaminsky:
had cannulas of estradiol in the mouse brain and over-ectimized female mice that were there for about a normal gestational period. modeling the rise in gonadal hormones, the rise in estrogen that would happen naturally in women who are pregnant. And so we’re able to do some epigenetic profiling on tiling arrays to really ask, where does estrogen change DNA methylation in the brain? We know estrogen receptors are steroid hormone receptors. They recruit all these epigenetic modifiers, histone, methyl transferases, and acetylases, et cetera. reason to believe that estrogen would change DNA methylation. We saw big changes. Then what we did was in collaboration with my clinical collaborator and friend, Dr. Jen Payne, who used to run the Women’s Mood Disorder Center at Johns Hopkins and is now doing that at University of Virginia. We looked in women in the blood when estrogen was high. And what we found was a really interesting relationship. At the loci where estrogen changed DNA methylation in the brain, there was more epigenetic change in the women who were gonna get postpartum depression. So
Hannah Went:
Mm-hmm.
Zach Kaminsky:
she had a prospective study who was going to get depressed and who wasn’t, but we had blood from third trimester. So, you know, there was this correlation suggesting that women who were going to get postpartum depression were more sensitive to estrogen epigenetically. And so then
Hannah Went:
Mm-hmm.
Zach Kaminsky:
from
Hannah Went:
Mm-hmm.
Zach Kaminsky:
there, we just followed that with, oh, we should be able to leverage this. And we were able to use some machine learning methods to boil this down to a set of loci at TTC9b and Hb1bb3, risk with about 80% accuracy. These genes are pretty interesting. So HP1, BP3, you know, we know it’s involved in cognition. At the time we didn’t know, but like I said, it’s been 10 years
Hannah Went:
Mm-hmm.
Zach Kaminsky:
and other people have done interesting work. So they’ve knocked it out and they’ve knocked it out in pregnant mouse mothers. And when they knock it
Hannah Went:
Mm-hmm.
Zach Kaminsky:
out, the pups die because the mom
Hannah Went:
Oh, wow.
Zach Kaminsky:
stopped taking care of them, which is really interesting for a mouse model
Hannah Went:
Mm-hmm.
Zach Kaminsky:
of to stop taking care of the pups. And so we also know with TTC9B there’s been less work done, but we know one of its close homologs, TTC9A. If it’s knocked out and then you give estradiol supplement, supplement with estradiol, the mice become very anxious and there’s changes in serotonin signaling in the dorsal profit. So
Hannah Went:
you
Zach Kaminsky:
it seems to make mechanistic sense. If we think about, go back to Hb1bp3 for a second, one of the things that we haven’t published on yet, NIH to do a prospective neuroimaging study. So we’ve got functional
Hannah Went:
Mm-hmm.
Zach Kaminsky:
MRI at two and six weeks postpartum. And
Hannah Went:
you
Zach Kaminsky:
what we see is that the biomarker levels associate with changes in functional connectivity in areas of the brain associated with maternal response to infant Q. So
Hannah Went:
Mm-hmm.
Zach Kaminsky:
what does that mean? It just means that if your baby’s crying and you don’t have postpartum depression, you respond to that and it’s healthy for the baby to do so. part of depression can result in a failure to sort of clue into those cues and maybe not act, which in turn becomes very epigenetically damaging to the baby.
Hannah Went:
Mm-hmm.
Zach Kaminsky:
We know
Hannah Went:
Mm-hmm.
Zach Kaminsky:
that this can change the epigenome and lead to later life psychiatric stress-related comorbidities. And so we think that the biomarkers are really interesting in the fact that they’re basically potentially changing the functional connectivity, or at least associated, they’re just marking, just marking
Hannah Went:
Thank you. Thank you.
Zach Kaminsky:
these areas of the brain. We can’t say they’re changing. But I think it’s an interesting example of going from mouse brain to
Hannah Went:
you
Zach Kaminsky:
human blood and then back to the brain. And I should, I’m gonna stop here, not stop, but I’m gonna say
Hannah Went:
Yeah.
Zach Kaminsky:
real quick, I think it’s important when doing epigenetic studies not to stop at those case control associations. to have, you know, other endophenotypes, be they neuroimaging endophenotypes in your tissue of interest, or be they epigenomes in the brain, or even just cortisol levels, if you know cortisol is important for the disease. If you have a set of 500 loci that come from your first EWAS, run it through, do an EWAS against those endophenotypes, and take the ones that cross-reference it. because that’s how we found replicating loci. These loci
Hannah Went:
Yeah.
Zach Kaminsky:
is 80%. It replicates in over six cohorts now. So, you
Hannah Went:
Mm-hmm.
Zach Kaminsky:
know, just like our SCOTU finding has a number of independent papers, we’ve been independently replicated for these, and we’ve done a lot of replication work ourselves. And so how do you find replicating loci? That’s how I do it. I cross-reference with
Hannah Went:
Yeah,
Zach Kaminsky:
endophenotypes.
Hannah Went:
yeah, I think the reliability, especially with these epigenetics, because it’s so new, we have to do it. It’s kind of something that people glaze over and just hope like, you know about it, right? But I think it’s absolutely needed. So I appreciate you bringing that point up. And then, yeah, I’m obviously obsessed with epigenetics. I love it. I’m very biased. I think it’s a great biomarker. But I will be the first to tell you that you absolutely need some other type of phenotypic marker or outcome to measure alongside of it. Obviously, the more have, the better it becomes. If you have other omic data as well, like the metabolome data, transcriptome, phenome, the phenotypic kind of correlation, then that’s great. We just need to learn more about the correlation, whether it be correlation, causation. We can get into that conversation as well. So I’m glad you paused and brought up that point. So yeah, the work you’re doing in the postpartum depression is just fascinating. What about I know we talked about the app you made for Twitter for the suicide and kind of using machine learning. Do you imagine that being commercialized with the work you’re doing there or what do you think that would look like?
Zach Kaminsky:
Absolutely, yeah. So, you know, I think now with a decade’s worth of replication, this just keeps working, you know, and we are hoping to be able to really combine the digital and the epigenetic technologies to be able to predict risk in women. So, you know, predicting risk using the epigenetic biomarkers basically entails taken during pregnancy.
Hannah Went:
Mm-hmm.
Zach Kaminsky:
And then we can prognosticate risk using the models that we’ve created over the years.
Hannah Went:
Mm-hmm.
Zach Kaminsky:
But we can potentially pair this with a digital signature. And this is what our startup company, Dionysus Digital Health, is hoping to do, is to have an app that can guess at risk based on social media using a digital signature and then potentially the epigenetic test to folks to confirm
Hannah Went:
Mmm.
Zach Kaminsky:
risk. And I think the digital elements are of course cheap to run and quick, but I think there’s value in having an epigenetic test beyond that. And that is
Hannah Went:
Mm-hmm.
Zach Kaminsky:
psychiatric disease is stigmatized, especially postpartum depression.
Hannah Went:
Mm-hmm.
Zach Kaminsky:
It’s one of those diseases that, the main problem with it is that it gets missed. So it gets missed by doctors They may or may not screen consistently,
Hannah Went:
Thanks for watching!
Zach Kaminsky:
not all doctors, but certain places. And women don’t seek treatment. They may not recognize symptoms and they may not seek treatment. We know that only 40% of postpartum depression cases tend to be found, and I think of that as only like 20% that are actually
Hannah Went:
Wow.
Zach Kaminsky:
treated, right?
Hannah Went:
Yeah.
Zach Kaminsky:
Yet the rates are 13 to 20% of the population. new cases of postpartum depression every year in the US. And so, what, only 20% of that is getting treated? So, what if you
Hannah Went:
Mm-hmm.
Zach Kaminsky:
had a blood test that said, hey, watch out for this, right? Would you think, oh, I’m just not sleeping, or oh, my partner is just not sleeping, or hey, didn’t that test say you were gonna be at risk? You should
Hannah Went:
Mm-hmm.
Zach Kaminsky:
watch out for this. Could this be depression? Go to your OB at your six week post-giving birth screen, asked to be screened, right? This
Hannah Went:
Yeah.
Zach Kaminsky:
is what’s not happening, but we think that a blood test could prompt this. And I should also add that companies are coming out with new postpartum depression treatments that are allopregnanolone analogs that are really interesting, like Brexanolone is on the market, which is an infusion, and Ziranalone, I believe, on the frontier, it’s gonna be coming. Sage Therapeutics, I believe, makes these. So there’s gonna be new, new therapeutics that could be paired with understanding if you’re at risk. Whether or not you can take those beforehand to
Hannah Went:
Mm-hmm.
Zach Kaminsky:
never get depression or you just take them when you get depression, you know, I can’t speak to that at this
Hannah Went:
Yeah.
Zach Kaminsky:
point. But there’s lots of potential. So, we’re excited.
Hannah Went:
Yeah, yeah, I’m excited too. I’m excited for you all. And again, to keep following along, I think we’ll just see that part of the space boom with pregnancy. It already really is in fertility, and then the pregnancy, and then after pregnancy as well. So I think we’ll start to see a lot of those hopefully assets and resources be available. I think that the digital screening is really great because it’s a cheap way, like you mentioned, to maybe screen for people who may be more wanting to get this test or fit that criteria. finger prick or if they’re already taking your blood, just pipette that onto the blood spot card, send it in, you get results back in a couple weeks, take an intervention therapeutic. I think that sounds pretty great and something we need to focus on more. It all comes back to the preventative approach. I think I say that like in every single episode. But
Zach Kaminsky:
Absolutely.
Hannah Went:
that’s why we’re here.
Zach Kaminsky:
Absolutely.
Hannah Went:
So yeah, we’re getting to the end of this podcast. Just a couple more questions left for you, Zach. What are you most excited about now? I know we talked about some of your current research Is there anything that you wake up and you think of and you can’t wait to get on your computer and start working?
Zach Kaminsky:
I mean, I am really excited about this postpartum depression direction, I think.
Hannah Went:
Mm-hmm.
Zach Kaminsky:
I’m excited to, you know, so I mentioned earlier that I love coding, you know, really just I love doing analysis and seeing what we can do
Hannah Went:
Mm-hmm. Yeah.
Zach Kaminsky:
with data. So any time I have like a new set of data, that’s, you know, I pretty much drop everything else and just delve into that. Even if I should be, I procrastinate work with with data analysis,
Hannah Went:
Yeah. I’m sorry. I’m sorry.
Zach Kaminsky:
which is technically work. But Yeah.
Hannah Went:
I was gonna say, it’s not that bad, right? You’re still doing something, but maybe something you like to do a little bit more. I have no coding experience. If I could go back to school, I think that’s what I would do is like bioinformatics and kind of learn that now. But yeah, I absolutely wanna learn it. Talk with ChatGPT, maybe they can teach me or take a couple of those free classes that are available.
Zach Kaminsky:
Mm-hmm. Yeah. So, I mean, it’s really, you know, encoding. We have one interesting thing that is unpublished that, you know, I mentioned, you know, we can predict suicidal thought from Twitter data. We’ve started to look at the responses and the other people
Hannah Went:
Mm-hmm.
Zach Kaminsky:
that respond to them on Twitter and see if we can score based on the responses, whether people will get better
Hannah Went:
Hmm.
Zach Kaminsky:
or worse. And we found that we can.
Hannah Went:
Oh.
Zach Kaminsky:
So, you know, then you could imagine, oh, well, can we make a tool that tells, sort of like, you know, with a chat GPT sort of help tells you what to say, what sort of response might be better, and can we leverage their own social
Hannah Went:
Yeah.
Zach Kaminsky:
media data to personalize that. So like this person, you need to talk about some of their loves, like recording music, right? Maybe that’s what they’re into, based on from what
Hannah Went:
Right.
Zach Kaminsky:
we can glean.
Hannah Went:
Yeah,
Zach Kaminsky:
So
Hannah Went:
that’s cool.
Zach Kaminsky:
I’m excited about that, but there’s never enough hours in the day, it’s, you know,
Hannah Went:
Yeah.
Zach Kaminsky:
to really bang all these things out. But I’m excited.
Hannah Went:
Yeah. Oh cool. I, yeah, I hear you on that as well. That’d be, yeah, if you can reply based on their previous tweets and their likes and their interests or even what’s in their bio or you know, different links. Um, I, I think that’s, that’s really fascinating. Well, um, yeah, you know, we, we’ve really come to the end of this, this amazing podcast. I have one last question for you. I ask everyone this at the end of the podcast, Zach, if you could be any animal in the world, what would you be and why?
Zach Kaminsky:
Oh, my 11 year old asks me this a lot. I should have,
Hannah Went:
Really?
Zach Kaminsky:
yes, but it always changes. If I could be one
Hannah Went:
Eww.
Zach Kaminsky:
animal, what would I be? Feel like there’s a lot of
Hannah Went:
Yeah,
Zach Kaminsky:
pressure on this question.
Hannah Went:
what are you
Zach Kaminsky:
I was
Hannah Went:
feeling
Zach Kaminsky:
so confident
Hannah Went:
like
Zach Kaminsky:
with
Hannah Went:
today?
Zach Kaminsky:
all the other questions, but today,
Hannah Went:
This
Zach Kaminsky:
this
Hannah Went:
one
Zach Kaminsky:
is
Hannah Went:
stumps
Zach Kaminsky:
the one that
Hannah Went:
ya.
Zach Kaminsky:
stumps me. What animal would I be? Even though I get asked a lot. I think for my son’s sake, I’ll say a cat. much a cat person. You know, going a little solo, a little doing my own thing. Also like to chill out and you know,
Hannah Went:
Yeah.
Zach Kaminsky:
I don’t know. And
Hannah Went:
That’s a good one.
Zach Kaminsky:
yeah.
Hannah Went:
No, I like it. You have to stand by it too. Don’t look back when you listen again and say, oh, I wish I would have said this. You stand by that cat.
Zach Kaminsky:
That’s right. The podcast went great except for the animal question. No, I’m standing by
Hannah Went:
Yeah.
Zach Kaminsky:
cat. Definitely. 100% cat.
Hannah Went:
Good, good. Well, no, like I said, this has been great. I’ve learned so much myself and look forward to kind of going back through your papers with now the knowledge that I’ve gained. So for any listeners who want to learn a little bit more about you, where can they find you? Web page, Twitter, I don’t know if you’re on any of that.
Zach Kaminsky:
Yeah, I mean, I don’t really use, as much as I use social media in my research, I’m not
Hannah Went:
Yeah.
Zach Kaminsky:
really great about posting on it. You know, I have a, I’m sure the Institute of Mental Health Research at the University of Ottawa’s, at the Royal is, probably has a lot of information on me. That would be good.
Hannah Went:
Perfect.
Zach Kaminsky:
So, yeah, I would go there as a first start. And that’s the email I tend
Hannah Went:
Cool.
Zach Kaminsky:
to answer too. So, yeah.
Hannah Went:
Yeah. Yeah. We’ll point point people in that direction if they have any questions or want to learn a little bit more. So, uh, thanks everyone for listening in and joining me at everything epigenetics podcast. Remember you have control over your DNA. So tune in next time to learn more. Thanks Zach.
Zach Kaminsky:
Thank you.
