 So we're going to continue with discussion for, for this session. Then we'll follow by a round table. We'll have lunch after right at lunch. If everyone, before we go to lunch, we'll take a quick group photo to share with the world. And then after lunch we'll return and then after lunch will be an opportunity for all of you to give us feedback as funders in terms of possible new directions of the field. I've made a spreadsheet for people going to the airport after meeting. So if people want to sign up at what time in which airport and look in your email boxes and it's just a Google sheet you can sign into so that people can sort of coordinate travel. Thanks. Okay, great. Okay, great. So welcome everybody. This is our panel discussion on families populations and societies. I'm happy to introduce our panel and I'll start with our bell heart pack. He's an assistant professor at the University of Texas Austin in health informatics data science and integrative biology. His research program as a focus recently on genotype by environment interactions and complications of what he calls sad effects stratification assortative mating and dynastic effects. Daphne Marchenko assistant professor and biomedical ethics at Stanford. She's an expert on the complexities of working on sensitive social traits the ethical dimensions of studies on complex traits in humans. Melinda Mills, who is joining us online from Oxford, where she is a director of lever humor Center for demographic science. She's been focused recently on biobank scale studies of complex traits, given a lot of thought to gene environment correlations in particular in the context of some of the discussions we'll have here today. And then Carl Veller, who is standing in for Alex young and is an assistant professor at the Department of ecology and evolution at the University of Chicago. And Carl is a theoretician who recently has been doing work on understanding family based designs and some of the complexities of those design strategies. So, to facilitate the discussion we're going to go back and forth between the panel in the room and try to facilitate some, you know, whole room discussion so I'll try to queue a sort of time for new questions versus opening up to the floor for discussion. And there are several questions here and the first one is queued up nicely by some of the studies we heard this morning so Loek in particular mentioned in his talk how family studies may be a productive future avenue for research. And so there are big questions in terms of, you know, thinking in the future oriented perspective of this meeting about family based study designs and in what ways are they useful for illuminating genetic architecture of traits. Will they get us fully out of the challenges of confounding and stratification. And for what kinds of problems are population based uos is still a useful way forward. So we'll start on that theme with the panelists and then take comments from the floor. All right, so does anybody want to jump in here first panel, maybe Carl you want to start. Yeah, yeah, so I'd like to touch on something that came up in Loek's keynote and in Peter Fisher's question. The mic closer and speak louder. Yeah, so I'd like to touch on something that came up in Loek's keynote and in Peter Fisher's question, and which I think deserves further discussion in the context of the topic of the session and that's the difference between the sort of standard traditional population based association study and a family association study. So I think we all appreciate that family studies can control for much of not all of but much of the genetic and environmental confounding and family indirect effects that are a problem for population based association studies, and which I think we're also a little more circumspect about our ability to control for now. So I think we all recognize that family based studies are indispensable in estimating direct genetic effects. But I think they also come with a bunch of issues that are less well appreciated. So, in some of the debate around Paige Harden's recent book, Molly Chavorsky and Graham Coop argued that perhaps the growing focus on family based studies is moving us away from the sort of basic questions that motivated our interest in GWAS in the first place. And I think that's very provocative. I think it's true in a bunch of ways. Some are relatively straightforward. I think family studies are influenced or some study designs of family studies are influenced by sibling indirect effects in a way that population studies aren't. I think sample ascertainment has different biases in family based studies versus population based studies. And I think maybe more subtly in the presence of gene by environment interactions about which we heard a lot yesterday and today. The fact that family based studies are sampling a different subset of the genetic variation relative to population based studies means that potentially they're subsampling a different distribution of environments or genetic backgrounds and therefore a different distribution of effects. And so, I think we need to recognize of course that there's some things that family based studies can't measure but we also need to do a bunch more work to try to understand precisely what family based studies are measuring. Thanks Carl. Melinda, do you want to chime in on this question. Sure. Yeah, thank you. You know, I broadly agree with with what you were saying but I think it also has to do with the trade as well right so whether a family based design is useful or not. And I think our bell was going to talk about, you know that the big problem with G was which we all know the population stratification and assorted of mating and these indirect parental genetic effects and then I think they can be really useful. You know, especially, you know, going really practically thinking about if you could do estimate these trio G was as with offspring and spouses, or some of the sibling models. You know, like the trio models they can they can show and they can show you what amount of genetic confounding you have and I think, you know, it's really been shown and I think that's in the next panel that we're going to talk about this that you're going to have issues and it will be more important or more relevant for the more complex behavioral traits, which Abdul was talking about as well. And I agree about the family models and then that's been a focus by as as Bogdan and Jen and some of the other speakers spoke about I think actually more diversity in samples across geography. Across life course stages and ages and socioeconomic groups, I think that's probably the stronger way forward and I can all stop now because I know you want to have more discussion but studies like our creature health the new 5 million study that that I'm involved in and many of us are in the UK. Over sampling low socioeconomic groups and different ethnicities it's important, but it's a different migration history in the as well to from the US and other countries. So yeah you get more diversity but completely different. Not relevant, according to the country. Great. Thanks Melinda. Well, do you want to try them in as well. Thank you Melinda and call and john. So, you know I've been thinking a lot about what I have to contribute I feel like most of my areas of expertise are much better covered by folks in the room. So I'm glad that you defined sadness is my expertise, because I do have expertise and vast experience. So, you know, I want to zoom out just a little bit before that and give sort of my, my two cents on a general theme of the meeting as I see it, you know, genetic architecture and how that depends on context. So I think, you know, in order to also think about where the money should go we should start by defining some of the holy grails of this question. You know, to me, you know, while important like you know it's not a holy grail to to improve our square by by 3% and so, you know, I think, you know, a holy grail to suggest is, or sort of, you know, something that we should champion as geneticists is mechanism. And that's a sense in which we should care about context dependency. And of course, you know, strive towards a generalizable answers and, you know, and I think in that respect, what we're, we're missing a lot is theory of how genetics depend genetic effects depend on environment and depend on context. And, and, and I think that's true for both the molecular and cellular level and the the environmental exposure and social level. And, and, you know, something else that that that I think we're missing alongside theory is pretty strong case studies that we can start and generalize from. And in that respect, I think, you know, some of the work that was discussed yesterday from Francesca Luca from from NASA, see not Armstrong from from from Jonathan's Jonathan pre shirts lab is, you know, it's really great for steps in that direction. Yeah, I think I'll. I'm much more interested to hear what Daphne has to add. Just a couple, hopefully you can hear me and this is on just a couple things that I wanted to add one is which I think Molly correct me if I'm wrong but you and Graham had a pre print that came out like yesterday on within family confounding and Carl's on it to awesome. Okay, so I just wanted to shout that out. But the other thing that I was going to say which I think picks up on what Jen had mentioned in her presentation when she talked about reflecting on the questions that we ask who we include in the systems that we model. And I think another thing to add on that in addition to thinking about you know what context are being captured and what are not what people are we including and what people are we not including. Also thinking about and I think maybe this touches on some of what Abdul mentioned in his talk but what are the downstream implications of the data that we generate and produce. I think sometimes it can be difficult to bring back into the forefront of our minds the fact that the data that we work with in genomics come from humans and thinking about what are the implications of this work that we're doing for those humans that have given their data in the first place and I think you know we heard in Bogdan's talk reflecting on well you know if we're implementing polygenic scores without calibration there are and I know Alicia also talked about this their health equity implications for this so just as my bioethics hat on in the room here encouraging us all to also think about the downstream implications of the work that we do and the human beings that will be affected by these studies that we've been hearing from over the course of the last two days. Thank you. Okay so on this topic of family based designs follow up questions comments from the floor in terms of the advantages disadvantages of these designs. I have a follow up question and Carl's comment about family based designs potentially sampling a subset of the context is that largely from a kind of ascertainment bias of the, inevitably having. Well, you have a sub sample of your large cohort that might have enrolled as families and that being sort of biased and who you know which families fully enroll versus the larger cohort or is there something more subtle just wanted you to expand on that. That would certainly be an example. What I had in mind there is that is that. So the genetic variation that's up sampled in families specifically specifically families a sub sample that can produce within family genetic variation because that's what within family association studies and other genetic studies are using. And that's and so that's sort of up wait certain families relative to others. And if those families are living in environments that interact with genetic effects. Then by up waiting those families you're you're up waiting your sampling of those potentially unusual genetic effects and so the the overall estimate that you get out maybe doesn't reflect the population as a whole. And wouldn't be what you'd get if you you know in like the perfect world if you had a an unconfounded population study it's not what you'd estimate from that study. And then one more clarifying question the up waiting of those families how is that happening is that it because of the need for heterosexuality and the parents are not exactly that it's exactly that so I mean if you take a if you take an association study at a single locus. And you perform that in the population as a whole and sample taken from the population as a whole. Then you essentially see all of the genetic variation at that locus and you try and associate that with trade variation in the sample. But if you're only looking within families you require that to be within family genetic variation and that can only be provided by parents who are heterozygous at the locus in question and those families might have an unusual distribution across environments. Great. We will all have to read more in this preprint that just came out. Okay, so any more comments from the floor on family based designs. Yeah, Peter. Just a general comment for discussion really so if you listen to yesterday and then today one conclusion one might draw is that G by E is ubiquitous in for human traits. But what's the what's the actual primary evidence for that it's it. You know if we look at the specific context where you can where people have looked at quantifying interactions like a genetic correlation for example between traits in males and females at the population level or age young and older people people have done that. Yes, you find some but by and large these correlations across these contacts are very very large. And so my question is what's the what's the primary evidence there's interactions rather than there's environmental effects which are like, you know that we know that we normally try to adjust for where you do a g was like effective, you know, sex of age of where you live or which snipper a use or whatever that kind of stuff how much how much interaction is there real interaction. interaction. Melinda, do you want to go or Arbel, I think you could be natural on this. I can't Arbel you're really small on my screen so if you're nodding if you want to go first or otherwise. Shall I go. I'm only tiny bit bigger here but yeah you can go. Go ahead. Yeah, so yeah, Peter it's a really good point I mean and it's, I think Arbel is also talking about we need to think deeply in deeper terms, theoretically, but I also think we have to speak in deeper terms in statistical thinking about this too. So we pull out that evidence of interactions and, and you know how can we piece out what's environmental effects or not and I saw if I sound like a repeating myself but I think it really differs in terms of like whether it's a complex behavioral trade or not and that and then the family models become more important in that way as well too and I, I think, you know, we're thinking we're using polygenic scores now to study sort of moderating and mediating roles of environment and genetic confounding. And then I think you run into a lot of issues and I think we just have to think a lot in deeper terms about, you know, and I've wrote about thought about this before about problems of conditioning on a collider and collider bias and spurious associations and you can really run into a lot of issues so I think we really need to scrutinize this more for statistical figure out what's the relevance of the polygenic score in the covariates and do you have endogenous selection bias. You know, what are the, what's all of this hidden heterogeneity because I think there's a lot of things underlying the relationship between the covariates with the outcomes and the G by E interdependence and the G by So, yeah, I think there's a lot more that has to be done in statistical terms too. Great. Okay. Yeah. So I guess, you know, the question can be addressed both in terms of sort of, you know, priors in the absence of data and then you know what data we need and in terms of priors in the absence of data I guess. You know, I don't know why we should expect the effects, the genetic effects, causal genetic effects to be different between two different ancestry groups when you know they're largely the same as in mouse. And, you know, on the other hand, we have, you know, all the evidence in the world that and, you know, a million billion dollar industry and plans based on G by E being being a thing in virtually any other environment where we can control and manipulate environments. So, you know, that's where my prior comes from and that's before kind of considering humans also being a little more complex than than plants in some ways. In terms of data, I think, you know, you know, hardly my prior or anyone else should prevent us from, from looking for those case studies and starting from there. And, you know, and I think in the population level, you know, I think we're going to discuss some of the diversification of GWAS in later on, but, you know, I think all of the conversation about that has been largely guided by diversifying GWAS in terms of genetic or just in terms of phrase actually. And, and I'm unaware of large efforts that are guided by diversifying context with that, you know, environmental or social context. So, so I think that data is largely missing and, and also really, you know, thinking about this at the mechanistic level of, you know, so questions that some very deep questions that came up in the discussion yesterday is include, you know, our environmental or context or environmental effects are they funneled through the same pathways as genetic effects. And, you know, if they are, does it, you know, make sense to start with and with a model of additivity of environment and genetics. So, you know, I think I think we're lacking on both data and theory. Bruce. Thanks, Bruce Walsh. Your comment about plants brings up an important point. We know a lot about G by implants because we can therefore detect it. The other thing is if your environment sucks, environment's terrible, it's not going to do the deal with it. If you're an animal and the environment's bad on a micro-environmental scale, you deal with the behavior. So one of the interesting features is a lot of G by E is handled in animals and mitigated by behavior. And so when we talk about the G by E of something that mitigates G by E, it's another way to think about something on that. In animals, we see G by E but only under extreme conditions. If you look at the milk yield of cows in New York State and in Mexico, it's the same genetic ranking. You have to move them to Thailand where they're suffering from severe heat stress to get a decent G by E, a re-ranking of the genotypes. So the evidence from domestic animals is that the G by E's are only important if you make a massive change in the environment. Okay, so maybe I'll use that as a segue. Well, let's say I want to make sure this, we move to some other subjects. It's kind of been fun about this meeting that it's become the genetic architecture of G by E or sort of so heavily emphasized. But one of the questions I think is important on the table. Of course, this is not related to G by E of course, but in fact, but the question of diverse sampling and and as we look towards more samples. This question that's sort of posed from from the accumulation of surveys ahead of time and it's sort of in the room as well. So in constructing more diverse cohorts, should we be, you know, the emphasis on sampling more diversity in terms of genetic variation versus environmental variation. And it's a segue from your comment in terms of is some of this because we are not, we don't have the variables to stratify by environments. We don't have the environmental cohorts that are diverse enough in those dimensions. So in that regard, yeah. Yeah, I just want to extend the question so a tour Bell, Sorbell started with the prior but I wonder if you could comment about the posterior from human studies about the what we know about the importance of G by E. Or at least the current posterior. So I can give an, you know, perhaps talk about this isn't within the example of sex dependent genetic effects, where we've, we've tried to develop a new way of fusing a GWAS data to, to look at how genetic effects differ between males and females as, you know, the estimate of interest and do that, you know, in a in a polygenic framework. And, and, and what what we see is, is oftentimes at odds with what the literature before has shown. And I think, you know, part of that is, is really just like, you know, trying to shift the methodology to thinking about complex traits and and how and you know, somewhat inspired by by the work that that has been done following the the Omnigenic model and what the Omnigenic model proposes which you know one of the the the biggest takeaways for me at least was was that we do need to to look at it in a different way and, you know, the generalization cannot be done from from single from side by side looking, you know, side by side and generally from that but actually going down a level and thinking about, you know, systems systems biology, thinking about about sort of, you know, gene interaction network to to better understand the GWAS data as well. But, yeah, and so what we do see lots of gene bisects interacting for example in in traits that were previously considered to to not be sex dependent, because of, you know, sort of like perfect or near perfect genetic correlation between males and females or because of high concordance of GWAS hits. So we just think that that's not a complete way of understanding interactions. But, you know, if I could even like, you know, give give up the whole retrospective study of the, you know, like wonderful UK Biobank and get, you know, different data from early development. Or, or, or, or IPC, you know, and to ask that question there, then I would, you know, do that in a second. Can I add something on this point. Yes. No one's saying no. So, yeah, no, I just wanted to talk about that that sort of the massive changes in environment as well that was raised before in relation to animal versus human studies. I think, you know, this is where where it gets really tricky and we keep having problems with this with this study of gene environment interaction because you need this huge exogenous shock for humans. So, Abdul was presenting the work about Estonia so when, when you had the Russian withdrawal they completely redesigned their entire education system. So you'll have a huge shock or you'll have economic shocks or you'll have changes in educational policy reforms or all of these different things. So you can model it in some way where you have some sort of, you know, exogenous effect but I think it's really hard if you think about it like modeling it like an instrumental variable it will always be really hard to isolate. It will always be related to multiple effects and I think that's the difficulty with some of these, you know, gene environment interactions you need huge samples in order to detect any sort of effect. And it's really hard to say okay well is that you know, can we isolate that causal effect that policy that change that environmental factor and I think the nice example was given by Jen about pollution, you know, all of these different contexts and it just has a very different contextual aspect, but I think, you know, one promise could be looking at these really extreme exogenous events or different contextual factors and that's been fairly successful in some studies. Magnus. Okay, Magnus has his hand up for a while actually. Okay. Magnus deferring. Okay, so go ahead. So, I want to follow up on that. I think one thing that just came up with the plant examples like well and plants will have huge variation there's like super high droughts or plants can't move but there's like if you ask a demographer most people can't move either. Right, most people's housing isn't as flexible that they can actually move so that's sort of a bit of a misconception that people move a lot. And also if we want these things to work I think Melinda gave sort of a start here by being a very sociologist about it giving the sociological answer is maybe we're not the people to decide what environments to measure. Right, I mean, we're here grappling with the idea of measuring environments but none of us are well few of us are social scientists or the kind of people that actually do this for their day job right. So, are we the right people to do that part of the G by E, should we be deciding what the E is in the G by E or should we just be bringing the G. That's a good segue I wanted to set up Daphne for a question on the this aspect of interacting with social scientists and the role that social scientists might play in this kind of research going forward. You might have noticed in the RFA that's been posted. The R01 includes a plan for transdisciplinary interaction and so maybe Daphne if you want to speak on those issues it'd be great. Yeah, well I was going to say I think it was Francesca yesterday who said that you know we don't need to reinvent the wheel and I think that that speaks to exactly what you're saying Michelle about there are other people outside of this room who are experts in the environment and and figuring out what contexts matter and how to measure the environment and so I think that also picks up on what you're saying John about. There's been a lot of conversation about interdisciplinarity and transdisciplinarity and you know the NHGRI strategic vision talks about this NHGRI strategic vision also talks about doing more work on the ethical legal social implications of research into complex traits but there's not a lot of meaningful collaboration between these entities and so I think you know Jen was talking about incentives I think we need to recognize that the incentives that exist right now and the way that our funding mechanisms are structured really operate to have people who may have expertise in a social sciences or ethics related background but they're still working in their silo within the larger consortia project so I think that that's a kind of structural system level issue that's going to need to be grappled with how do we work to ensure that when we bring in social scientists when we bring in bioethics folks that they're not positioned as the assistant to the scientist supposed to help them with research communication or figure out you know what are the downstream implications of it. I do a lot of work trying to understand how genetics researchers and LC ethics people think about the goals and values of collaborations with each other and I see time and time again that oftentimes there's a disconnect between what bioethics folks or social scientists think of the goals of these collaborations being and what genome scientists think the goals of these collaborations being and I think part of that is driven by this incentive structure and what we value when it comes to tenure and promotion but if we are meaningfully wanting to do interdisciplinarity then we're going to have to change those systems I think. I think we have time for just one or two more comments. So let's do Magnus and then we'll approve both hands for a while. Okay Magnus is over here. Yeah I just want to push back slightly on the whole plants versus animals thing here so it's true that there's a lot of local adaptation in plants right but I mean in addition to working on plants and so card carrying evolutionary virologists right and there you know there's plenty of local adaptation in animals and sticklebacks for example enormous reaction you know move them from salt water to fresh water etc right there are lots of lots of effects there right so that's one point but the other thing I'm struck by when I listen to these discussions is I think it feels like the way environment is discussed there's something very simplistic in the sense of you know we think about it's how much fertilizer you give a plant or how much feed you give cattle or something like that right they're completely exogenous things but many of the trades we're looking here in you know where humans really do stand out is that they literally inherit their phenotype right from socio-economic status to language to a whole bunch of other things and this is where I think you one really needs to think slightly differently and question whether the models and frameworks that I applied make any sense at all. That's great. Yeah so just wanted to make two quick comments one was to echo what you the question John about sampling and diversity I was thinking this is actually very important I'm taking the context of some of the initiatives happening in Africa of new where we know the genetic diversity is such that we're going to struggle to account for population modifications and therefore we should be thinking as those initiatives are growing to essentially bring family sampling as soon as possible because that probably one of the best opportunity will have to deal with those issues. So the second point was about I guess what definitely just said and I think I usually when I enter this field as a sort of statistician basic statistician I've heard about the boogeyman population stratification and I think we should be embracing that boogeyman in some ways and the way to do that is to recognize that the stratification is a trait in a way. Abdel I think very nicely illustrated that those patterns that we see in those in those samples sometimes reflect ascertainment but this is a trait it is and you know migration this is a trait and I think the more we will spend time thinking about and embracing that as a trait and pushing to better understanding that I feel like that will probably ease our collaborations with social scientists. Okay in terms of time we're going to let one more comment from guy and then we're going to stop we're going to continue the conversation in the round table run by doc but since we had a longer late start to the morning we're shaving this off by 10 minutes and shaving the round table off by 10 minutes and we'll be continuing to discuss. Guy last comment on this panel before we get around well I just want to maybe emphasize a something that they are well said beyond beyond the need for better data is that I think that the quantitative kinetics way of thinking of this expansion of residuals in terms of dealing with G by E might be in underpowered to detect some types of environmental interactions just because they end up extending into deep parts of the expansion in a way that you're always underpowered to see them but actually finding better ways to model G by E and then analyze them like our Bellman this beautiful study about the effects of sex on traits by taking a different kind of framework actually revealed a lot of effects that couldn't have been seen before because of the modeling framework and I wouldn't be surprised if when when that is actually applied to humans where we have other independent evidence of how a social economic status and other factors usually influence traits then we would actually see it in the US type studies. I really do want to not cut off the round table so I think the point that guy just sent it on is sort of a good we've I think discussions surfaced a nice tension between these ways of modeling and thinking about G by E and I think that's maybe productive for us to think about going forward so let's thank the panel. So now we'll be moving on to our our round table discussion. Oh. This is on. Hello. All right let's do this. Okay, so this is the last round table topic is quantitative versus qualitative differences in the genetic architecture of biological and social traits. I'm Doc edge and I'll just, I'm going to ask the panelists to introduce themselves. Disciplinary background from which we're coming at this question and to make a comment on one of our three questions which aren't on the board yet but I'll just say the first big question is what is a social or behavioral trait, or is it a continuum is there some other factor that is the interesting thing that we should be keying in on. Are there important similarities and differences between behavioral traits and other kinds of traits. And the third question is, what do we need in terms of methods or conceptual frameworks or data to understand or intervene on on more biological or more social traits. So with that said, so Graham Coop was supposed to be on this panel. He wasn't able to be but he has a comment that I'll read after the panelists go through but I'll ask all the panelists to say a couple words. Jonathan, would you, would you mind starting. You mind starting me should we go the other way. Michelle, do you mind starting. I don't mind starting. So you want me to comment on one of these right and okay my background is as a psychologist. And a bit of an epidemiologist, and I'm taken, I was like inspired by NASA see not Armstrong's talk yesterday, which basically lays bare the issue that none of these traits are any of these things right because we had a snip that influenced a pathogen to your skin to a pathogen so that's ecology. So we did genetics biology ecology, and then I guess the spread of the pathogen in the tick that was it that's definitely, you know, an efficient wildlife control, whatever you want to do. And then whether people recognize that they're a bit by a tick that's like public health or information or education like the entire causal chain could be interrupted in any of our disciplines. And so it may be the easiest interruption is just like public health you just talk to people about tick bites and their, and their danger could also be that one of the other things is, is an easier intervention so that's just one snip effect on one one cancer. That that's transverses all like many of the disciplines, sort of that could feasibly relate to it so to think of the trade itself as any one of those things is just to me. Fairly weird. Sorry. Hi, I'm, I'm Alison Bellin. Sorry. Okay, thank you so much for the opportunity to be here I cannot tell you how much I enjoyed this workshop I have learned so much I've taken so many notes so I actually study animal behavior. I love plasticity, which I feel like might be a little bit of a bad word here, because we're talking about g by e but I actually haven't heard that much talk about plasticity. But I actually think that how and why and when phenotypic variation arises, especially behavioral variation is a super interesting rich question. And I like the messiness and so it's been really fun to be a part of this workshop. I think maybe the first point that I would want to speak to has to do with the distinction between social and biological and I'd like to actually just question that distinction a little bit, because I think that the social can be biological. And biological can also be social. And I wonder how much we gain from making that distinction, especially because it can lead us back into some nature nurture traps that are not particularly productive when we're actually trying to think about how phenotypic variation arises. There's really great examples about how social environments can influence biological processes. There's also really cool examples of course about how biology can influence social environments so genetic variation and habitat selection for example different genotypes actively seek out or experience different social environments that can influence their phenotypic development. So I think these lines are really blurry, but I don't think that's bad news from my perspective. I actually think that's really interesting and an opportunity to really understand where variation comes from. Thanks, Mike. I'm a quantitative geneticist who works mainly in domestic animals. I agree with Allison that there doesn't seem to me to be any hard distinction between biological traits and social traits. You have to worry about environmental effects in all of them. There might be more indirect effects in social traits. In livestock we've long recognized maternal effects, so the effect of the mother on the phenotype of the offspring and this can be partly genetic and partly environmental. There's no reason of course not to have a paternal effect as well in the model for the effect of the father and it could be partly driven by his genotype and partly by environment. More recently people have looked at what for a human audience I'll call housemate effects. For example, if you're a hen, the other hen's in your cage and it turns out that for some traits they have a large effect and you could study this. You could put housemates in the model, both their genotype and their non-genetic effects. It means that if your housemates are your brothers and sisters, so your genotype will be correlated with theirs, that when you do a GWAR you'll pick up both the direct effect on the target person but also the indirect effects from their sims or housemates. So that's an interesting model that's probably worth exploring in human data. Thanks, Aaron. Hi. My name is Aaron Panofsky and I am a sociologist but rather than a sociologist who applies genetic techniques in the study of sociological dynamics or variables, I'm a sociologist who studies scientists and how they think about and use various tools in their work and also how their tools are in broader social context. So I wrote a book about the field of behavioral genetics and its history over the last, you know, 60 or 70 years. And I guess I would take maybe a slightly contrary position to what's been taken so far in the panel, which is that I think that lots of fields, and this goes back to our questions about interdisciplinarity and what it would mean to collaborate, but many fields actually take very different definitions of behavior, very different definitions of traits, and I'm not sure they're all commensurable necessarily in the way that we have discussed. I mean, I guess what I would say in some sense is, though I agree that social versus biological there's, there's entanglements between them but I'm not sure that that that distinction or the traitification of behaviors or human attributes is necessarily exhausts what we think about behavior and so forth. Yesterday I asked a question about changes in drug addiction over time or drug response or even what drugs are available in a society over time and I think that the, the broader sort of cultural contexts are deeply implicated with what we think a behavior or what we think a trait might even be. So let me just give you two quick examples. One is that you know so in the for historians and cultural anthropologists. I have an idea. So this isn't a topic anyone's brought up today but I think it's a topic that GWAS has studied. So sexual orientation, right so there's a controversy over whether sexual orientation is actually a thing, right so this is a way that we conceptualize sexuality in our contemporary society. First of all there hasn't always been a category of homosexuality right that categories of relatively modern and recent invention, and it's not obvious when we look cross culturally that sexual orientation, right that there's sort of a deep that we each have a deep orientation towards certain sexual certain certain objects as a as an expression of sexuality and objects being other people usually that that is always true in all societies. So does it. So, so what is it so I guess that that just raises the question of what it means to do something like a genetic study of something like sexual orientation and whether we can sort of say that that generalizes and tells us something about humanity writ large as opposed to a certain slice or way of thinking about things in one in society. I think disability is a similar one right disability isn't just a, you know, you have some biological defect, but it is an entire sort of gestalt a whole way in which society deals with you. It's also a moving target right so I just filled out one of those forms, you know and I was like oh wait, I'm just, I didn't you know I never thought of myself as disabled but when I went down the list I was like oh I guess I have to check this box and you know for one of the things that is really mild. But, you know so these things change over time so I'll stop there. So my question my general question is just I'm not sure that that tradeification is a universal way to think about about human possibility or human differences. Hi everyone. I am a Gossiper. So I'm not from none of those disciplines. My background is I come from music and I have a bachelor's in journalism, a master's in anthropology with a specialty in evolutionary medicine, which I taught for 14 years. Then I went into public health and got a second master's in applied bio statistics and then I have a PhD in psychometrics. So I gossip about data and I measure shit. That's what I do. When thinking about this topic in the conversations I've been hearing over time. Let me give you a brief example. So what I do in the field of Alzheimer's now is try to track racism from across 100 years and map that onto Alzheimer's disease risk for populations that have greater risk like people racialized as black and people racialized as Latinx. And what we find in my studies is that racism, often before the pop, the individuals in the study were born is associated with their cognitive decline over time. Now why would that be. And that speaks to the question that was sort of raised in the previous panel, which is sort of the relationship of population health and family studies. It is impossible to tease apart. Inheritability in utero exposure early exposures from a legacy of racism in our entire world. And so before we get to a place where we want to better genetically capture the sort of in family differences, why don't we start seeing how these things cluster based on the facts based on structural sexism, not gender based on structural racism based on capitalism. Let's see if these things are getting underneath the skin. And then we can start to tease apart whether or not these things are inherited or not. Thanks, Eji. Jonathan. Hi, I'm Jonathan Pritchard. I'm a population geneticist and I'm filling in for Graham Coop and I can only imagine my main qualification for this is that I have similar accent to Graham. So I wanted to think a little bit about the kinds of goals of different sorts of studies, and like what do we want to use these types of approaches for and I think that that has a big impact on how we should think about it. I wanted to highlight three different areas. So, so one is how do we learn about biology from GWAS. A second is how do we do personalized prediction for medicine. And then a third is, how do we use genetics in other fields. So for example in behavior genetics and sociology and so on. And so I think that these really should drive us to think about, you know, different aspects of the design. So for the first one, for learning about biology, you know, I think we care a lot about learning about both general mechanisms of biology, as well as perhaps identifying specific genes that can be drugable targets. And so for this, I think that a lot of the things that we need to do are sort of more on the functional side. I think that I talked, I spoke briefly yesterday about how I think there's a huge need for understanding how, you know, sort of what are they mechanistic links between genetic variation on one end of the scale and, and, you know, disease on the other end of the scale. So how do effects flow through gene regulatory networks, how do they affect cellular phenotypes, how do they affect tissues and, you know, how does that flow through to organisms. So there's a, in many cases, sort of functional and lab based questions in my mind. I do think actually, this is a separate discussion, I think there's actually still a lot of space for collecting exome data, some really interesting work, for example, from Luca O'Connor showing that we can get different kinds of information from exomes and we can from GWAS. So anyway, so that's one set of questions where we want to learn about, about biology from from GWAS and, you know, that hopefully affects, you know, medical interventions and as well as our basic understanding. Secondly, you know, there's personalized prediction. And so we heard, for example, from Bogdan today and from Jen about, about PGS scores. And here I think that it's, it's clearly essential to, you know, think much more broadly about, you know, how, how individuals are situated in, in a specific environment, their, their communities, how these things intersect with race and genetically defined ancestry and many other factors. So, so I think that for that personalized genetics type of goal, we need to be much, much, much more comprehensive than we do on the biology kinds of questions. And, you know, so I thought the talks today really inspiring about how we might approach that and these are, these are really challenging problems. And then the use of genetics and, you know, in sociology and behavior genetics, that's probably the place where I'm least commented, least qualified to comment. But I, as far as I can see that, you know, that has much, maybe in some ways more in common with the, you know, what I was describing as the personalized genetics sorts of questions. But, you know, there I think we, you know, we really need to be in conversation with the people in this room and outside who are experts in those spaces to, you know, to think about, well, what are they actually the questions that they want to answer? And how, how does genetics intersect with that? And how, you know, how can we as a genetics community be useful to them? So, thank you. Great. So I'm, as I said, Graham couldn't be here, but I'm going to read out his comments. I'm not going to do the accent. I should have Jonathan wrote this. Graham was my postdoc advisor, so I could actually do it. So there are many axes that we could place traits along and the usefulness of those axes will depend on the questions we want to ask. I'm not convinced that the biological versus social access is a very useful one for, for many of the complex traits that we as geneticists study. Many human complex traits, at least those studied by geneticists will be influenced by the interaction of genetic and social environment played out through physiological developmental and behavioral processes that will be shaped by social interactions from early childhood onwards and will be subject to change as societies change. So dichotomy of GWAS traits as social or biology likely is unhelpful. I do think that a useful access for complex traits is to think about is the extent to which trait GWAS are confounded by genetic and environmental confounding in indirect effects. This falls within this discussion because one source of this confounding is due to traits shaped by social phenomena of how individuals, families, groups and societies organize themselves through a sort of meeting migration and the transmission of environments across generations. And two, the number of behavioral traits GWAS to a number of behavioral traits on GWAS are outliers in this confounding axis and show strong evidence of multiple sources of confounding being compounded together consistent with at least some behavioral traits being seriously confounded by social phenomena. This poses serious issues for the interpretation of these GWAS and raises question about the utility or at least the applications of GWAS approach for understanding these traits. While confounding itself might be an interesting source of questions about social processes, the fact that multiple sources of confounding are compounded together, combined with the distance between genotype and phenotype, make it not at all obvious to me that the field will try and traction on these problems. I'm very cheerful there from Graham. I have an interesting one that that basically observation that basically links to what Aaron was saying and also Graham, like, had you read out. And that yesterday when you said like, well, what if addiction is different seven years ago and and I'm a psychologist that's a super interesting question because like in psychology, sometimes addiction is viewed as very medicalized. Right. And so, for example, in the south of the Netherlands, people on average, pretty much white and of low SES, they're addicted to GHB, whereas in the US, they're addicted to different substances. Right. Is that even the same thing? And so, where I think GWAS can be very useful is if we let go of the idea of GWAS as anything like biology biology or anything, but is to GWAS addiction in the 70s and to GWAS addiction now or to GWAS addiction in the Netherlands and here and do a genetic correlation and we're not saying genetic correlation would reveal biology. I'm just saying, if it's low, that just points to the fact that those addictions aren't the same thing. They have different sets of causes. Right. And those and that they're just like, so it's more of a, there it becomes more in psychology. I think the next more becomes a medium to correlate things that are very far away. I would almost advocate dropping the word genetic from that correlation because it will confuse the hell out of people. Right. But the correlation doesn't, isn't caused as caused by genes in a very, very remote sense. But it's, it's a very useful indicator of things changing in our society. Right. So yesterday there was the example of statins changing who gets heart disease. But we know about statins. That's tractable. Many things in sociology and psychology in the fields we I sort of move around in, they aren't tractable at all. So it's, it's to me, it's like analytic gold to be able to go like actually the addiction wave you're studying there in the 80s for that certain substance. It actually, it doesn't share any of the top hits with my addiction one because it's a different substance. So maybe the nicotinic receptor gene has nothing to do with your drug. Right. But the genetic correlation is low or high, which points to like either shared or separate causes. So I would advocate, especially the social sciences. We can find uses for genetics, but it does mean we have to talk about, like, sort of what genetics then means that it just becomes a medium to correlate things that are very far apart in time, or in space. I think that's a really genuinely useful use of genetics. It's just not a common one. But it has important implications. Say we find that like the different ways of addictions that occur throughout time have different sets of causes. Right. That tells us something about interventions and then how to deal with. So that's sort of me responding to Aaron and some of the other things. I built on that. I think the irony of the conversations that I've been hearing for the last two days is that you've always only been using social traits. The fact that things were built on race or built on height were never genetic factors like there for me. I think a lot of anthropologists share this vision is that there's a lot that happens underneath the skin and the degree to which it needs to be a genomic or genetic level shift is greater than some of the conversations that we're having. I appreciate Melinda using the term shock in this because I often think that we're conflating everything underneath the skin together and using these genetic approaches to sort of talk about social issues through using words like ancestry, which is highly conditioned on social factors. And sometimes I resent the use of terms like self identified because I felt like it's really rude. Like you gave me a list of categories and told me to pick that. That's not self identified. I'm self identified as AJ. I am who the hell I am. You told me to choose a category. So that is already a social condition by which you forced me into and then you go pretend as if it's correlated with some kind of genetic or continental ancestry. So almost feel like we should step back and acknowledge that we're not talking about what we think we're talking about. It was all made up. I want to push on Michelle's point as well. And maybe channeling Graham's comments a bit here. If you do find that low genetic correlation between addiction in 1970 and addiction now, where you find a high one. Do you know that the causes are different? Or is this, you know, how do you know when the causes are different versus you're picking up some kind of shared structure? So if we believe sort of that the set of the total set of causes of an outcome, that's that's like society that's like a billion things, right, that I feed into something like addiction. And if, if most or some of those right are heritable, that heritability will propagate into it. And it doesn't even have to be that it's a heritable cause like if we were to, well, you know, if we if we in society sort of like select people on genotype, which we do all the time it's all kind of selection mechanisms and I'm not even saying a gene calls anything just could be society filtering on the gene. You know, it most apparent I think in beauty ideals, right. Society decides what it's a beauty ideal that's literally just, you know, society deciding what genotypes get to be represented in a magazine right so I'm not even saying it has to be a biological cause. But all those things will filter into your GWAS of the thing. And so if the set of causes were to stay the same, then then there's no reason to believe that like the genetic correlation between a thing in the 70s and now would change. Whereas if the if the set of things going into a changes, right, you'd expect the genetic correlation to go down. And so you aren't certain but there's many, many sort of fields of study in psychology that I'm familiar with where where we've been talking for ages about, you know, how psychiatrists in different countries perceive a certain disease how stigma is different about different symptoms and so is depression really the same in different countries. Maybe not even at the level of the psychiatrist maybe not even and those things cannot be like empirically resolved right and that doesn't always need to be sometimes just the observation is sufficient. But if you do a GWAS of depression in in two of those cultures where you have this inkling that actually they're very different constructs and people have done that and you find the genetic correlation of point for between depression in individuals of Asian ancestry and European ancestry. And I agree confounding a lot of things because the people that do the GWAS they're not thinking of social science right they're just combining every sample they can get from many different Asian cultures so it's not great sort of like way to do it. It's just as an illustration that correlation is point for it's way higher for other psychopathologies that that are full of to be less heterogeneous across a culture and context. And so I think that is information that is useful information I think that is sort of like actually something we can find a place for in social science. It does require redesigning some of the GWAS is with so with social scientists in mind right because you know, but anyway I think it's very it is a useful use of genetics and it's just a social use of genetics and it's not a. It doesn't even get to biology it's just like we're using it as an indicator at that point. So let me let me try and synthesize briefly and then and then go to questions from the audience so what I'm hearing is we had one access along which the discussion was supposed to proceed and and every panelist has as rejected that. The idea that there's a there's that this is the right access behavior versus social and multiple reasons for that you know one being that you always have to consider both environment genetics anyway so you know. What's the point and another being that you know all. So I think that those traits are in some form social as well because they're they're socially determined, even if they're under the skin. And I think this idea that Aaron brought up of tradeification I think is very interesting, because maybe there's a sense in which we find it easier to trade if I biological qualities than, than social ones. But that's all I want to say and I want to go to audience questions. Can I say something about that. Sure. Just as somebody who studies behavior I just have to say something about like how hard we work to try to make sure that we measure things reliably. And I totally grant that measurement is a really important question. So I studied non human animals and we do all sorts of things to make sure that what we're taking data on and how we're observing can be repeated that it's, it's repeatable it's replicable across observers over time etc. And so I always just get a little bit defensive like go behavioral traits they're so messy and you can't really study them because you can't really quantify them. I just want to push back a little bit that we that is something that we worry about a lot and have come up with methods to try to address. Can I just add to that. Sure. I agree. I agree completely with with what you said but and and really in some ways we're making life too difficult. A lot of this isn't so difficult. I studied fearfulness in both cattle and in dogs. Every dog owner that I've met who's got a free fearful dog says it's because somebody mistreated it. But in fact, the heritability of fearfulness is 50%. It's it's easy to measure. You can run selection programs and they work. There's no doubt it's genetic. They're also environmental effects. Nobody's denying that. But fearfulness is it's really no different from studying height or weight or something. You know, it's you have to measure it correctly. You have to put some effort into it. But you get you get answers by doing the actual empirical science, not by not by philosophizing about, you know, how complex it is. I want to like I want to I want to push back a bit and I'm going to like end up in Eric Turkimer corner, something I've never thought I'd ever do. But like if if if people treat dogs of a certain breed badly, they become fearful and that will be heritable. Right. So, like the causal mechanism causing heritability doesn't have to come from the biology because if we like all decide to treat the same type of drug. Very badly, and they all become fearful and you would do what she was you would find hits. You could even breed it out through a selection program because the dog we all sort of three badly would disappear. And there's like no way your experiment of my experiment looks so when you have maybe don't need philosophy if you want to just get an allele to go and like improve disease great but like in many other applications you do need to philosophy because it kind of does matter. Fundamentally, whether it's like true that the dogs are getting, you know, because the easier intervention than the breeding program is then to like, you know, maybe make sure people aren't being assholes to that kind of dog. I don't know. You kind of do need the philosophy. Okay, I want to go to the audience now we're having a good time up here. I have a question online from Neil Rich goes back to the trade application point. He says when large biobanks were constructed like the UK Biobank, some of these social factors like SES educational attainment or conceived as social risk factors or social determinants, they weren't considered as health outcomes themselves. So what I will point that the So the end of Sandra's question was at what point did the social determinants determinants become medicalized. Do you want me to get at that. I often think that the problem we have this just like, maybe people deleted history, or they think that what happens outside the lab or our work environment. Or some separate world we step into, but we're not different people when we cross those spaces and our history does not change. So that history of eugenics. So one of the things is things I track a lot of our historical movements into specific decisions made by institutions. So often track European colonization that we're living the effects of to great colonization of what becomes the Middle East Northern Africa and into parts of Asia, because it's that strategic movement of deciding people are different and then treating people differently. Based on this rule they created that then trickles down into all of our societies. That's the same thing when it comes to eugenics and science. We often push out to your example, the bad apple of eugenics and the way we do science and go that's rare. That's only a few people who did that. But that's actually the foundation of westernized science is to justify the social stratification that was occurring. So our science is never separate from it. It's always embedded in that and we're always trying to overcome it through our rigorous methods and models. But at some point we can't so I would say that it didn't become medicalized it started medicalized because we started producing anthropometric assessments of cranial size to justify who was more intelligent. We started studying black bodies and indigenous bodies to justify them being different than white bodies. We started there. That is our foundation. Are we going to decide to create a boundary between what is social and behavior now that is a choice that does not mean that it's true. I have a question to you to follow up. So I usually see sort of like and I've worked on social outcomes in us right I usually see well done social science genetics if it's really done well as a force to sort of this and basically to sort of like adjust perceptions about this like the within family that we have heard ability like so saying that's a lot of genetic gene environment correlation that work stems from doing a genetic study of education can really get to like all the things we learned about how if we do solid control their ability if those rates is far lower than than than we thought like say 20 years ago. So I was always thought like doing the social science genetics but doing it like really with a high fidelity and well is a solution but do you do you sort of what's your perspective on that. This kind of taps into something that was said earlier about case studies that our bull was getting at I think there's no better case study. Then knowing the history that European nations were in the dark ages in the 1300s and 1400s and poverty and starvation and through the act of colonization have become the better health of the entire world. There's no better case study for how social factors can shift in just a few generations the health of a whole population how you can go from self governance and indigenous societies with great health to the poorest health in a society such as ours. Some of us experience that just in our lifetimes you have a shift in the resources that you have your health entirely changes. And so I think we almost have like the case studies in front of us to demonstrate we just don't ask those questions because they become this sort of color blind we don't see them they just exist like that's just air. We don't need to name that that just is what it is what we need to name is why you're deviating from the norm wise your standard deviation far off from where we expect it to be. Why is why we compare you to everyone else why do you regress to the mean we're used to these sort of ways of asking questions. We have the case studies to get at what we really want to know about how the genes interact interact with our environment. Our bell is Mike. I think I think I. So I wanted to to be on some of the comments by the panel members and also linked to some of the comments made late in the previous panel is that, you know, I just feel remiss if I don't say this stratification is not a trait. Limit on our ability to do causal inference. And it's been shown in various ways to be a serious limitation on GWAS data in particular, or, you know, where, you know, correlation is not causation in the most in myriad ways and. You know, I'm if I'm even willing to go with, you know, Michelle and, you know, there's something to be said about or to be learned from a difference in genetic correlation between, you know, to societies or something of the sort you know I think there's, you know, I'm not convinced that that is the case. I'm much more convinced that there's a lot of harm done by. Just, just putting out there the genetic correlation between same sex behavior and risk taking from questionnaires to 23 me participants and. Or, or, you know, from from just, you know, proposing that people's genotypes make them move next to a highway or to a coal mining region where their risks of asthma or cardiovascular disease is higher. I think there's real evidence of the harm that we're doing with that and. And it's much less clear that we're learning anything about mechanisms including social mechanisms by using genetics, even if just as a proxy. We can go to Naomi in the back. I wanted to add to the discussion about the genetic correlation estimates that Michelle was talking about that I feel like there's not enough benchmarking when people do genetic correlation so. When we did the actually the first study of estimating genetic corrections from G was data which was across the psychiatric disorders I made sure that we did estimates of genetic correlations between data sets of the same disorder, as well as between the same disorder showed, you know, very high genetic correlation between the schizophrenia groups, lower between the bipolar groups and, you know, even lower between major depression. And so then when you go across ancestries, often people estimate a low genetic correlation and interpret it as an ancestry effect, whereas in fact it's a phenotype effect. And so, you know, just advocating for more benchmarking when people do these studies. I wasn't saying it was an ancestry effect. I was saying it was a phenotype effect like the phenotype and that's sort of like, this is the difficult thing when talking about genetics. It's like, I am interested in phenotype effects, like what why is a phenotype different in a different culture and is it different in a different culture or is it different over, you know, 20 years from now is a different 20 years back. And sort of I really respect your perspective on this and I do admit that there is sort of risk for misuse of all these data, right, because we speak freely now about scientific things but take them out of context and give a subtle spin on what I said and you risk instilling sort of like, oh, there's like genetic differences in the depression between these two ancestries. I wasn't speaking about ancestry and that's really difficult and hard to disentangle. Maybe in the limit we can't and then maybe we shouldn't do all of those things, right? So maybe there are things then we shouldn't study. That may be a reasonable outcome. I'm not entirely convinced that's the case. I do think people like in psychiatric research deeply struggle with intercultural differences in our diagnosis and I think those are genuinely important questions to ask. I think the same with like over time I am very curious to see whether psychiatric diagnosis 50 years ago would correlate the same with each other or how would they correlate to ours now if they morphed into different ones. Those are like genuinely interesting questions I think for psychiatrists to ask and you're right, there are phenotypes where that's harder to sell and there are situations where maybe people are naive to what their little number means outside the context and that's something we should definitely continue to evaluate and discuss. But I do feel by being open about that, like I think that social science genetics or psychological genetics can only really work if we de-essentialize it, right? And say like these are not biological facts. Genetic facts are not biological facts. They're just correlations as you said. They aren't causal and the correlations can be used in something in a different study but we have to do both of those things at the same time. Otherwise we're putting out numbers that like live in a different context that we intend them to live in and that is consequential. You're absolutely right. I almost feel like we have to make a decision about what we want to do about our findings and that's to my point about why we keep feeling like we're scientists during the day and then we walk out and we're an individual, a mom, dad, whatever the case may be at home. Reality is those findings have real implications for how we understand humanity and how we treat each other, how we structure our societies. So almost would rather you first make a declaration on what you're willing to, the places where you're allowing your science to blur with your political self and then do the science because then it won't be about protecting some way we've been doing it and it will be about how it actually provides some benefit. Because I'm thinking that example, if we determine that when you put people under these oppressive socioeconomic environments, they tend to, there's some kind of gene expression that allows for this behavioral trait of alcohol consumption or some other kind of drug to be used. Then what the implication is on you as the scientists then what are you willing to advocate for then changing your society so that we don't put people under those kind of conditions. Or is it just a finding you walk away from, you publish and you're done and that's not my job, that's someone else's job. So what if we complicated our findings in a different way? Jen? Oh, sorry, look, do you have a mic, Jen? Maybe we'll go and then we'll give the mic to Jen. It's great, great food for folks. I was trying to reconcile some of the points made by Aaron and the great point made by AJ as well. And I was reflecting on this stratification and I thought, okay, well, we can do anything that varies, right? If it varies, we can try to do as it. But then I want to ask a question to the panel, do you think that the fundamental problem is the variance and whether variance means inequality? Hallelujah. So that's for you. Anyone want to respond to that? I was trying to give other people a chance because look, now you're speaking my language. One of the problems we have when we match structural racism, so we have a measure for people racialized as black and a measure for people racialized as Latinx. And I have a couple colleagues that I'm building on other metrics for. It does not vary. Racism does not vary over time. When we use sort of, I'm not a fan of principle components, so it's just so disheartening to hear it repeatedly used over and over again is something that's so high stakes, like alleles, but you know, whatever. When we use our approaches, like a class analysis or some other kind of late modeling to look at differences, we find that it's often a shift at 1970 in our data where racism starts to look differently in this country. Now, I know my history so I know 1968 the Fair Housing Act occurs that allows for desegregation in our country to occur. So then we have to shift our model to look at racism differently after that time period. What if I mean in this almost piggyback so Michelle question like what if we shifted our statistical approaches based on our understanding of history, because it better maps on to what we expect. So then maybe variation is not the key. Maybe we are just comparing means or medians at a given time. I often say anything past 1990. I expect variation. I expect my country to have changed enough that there is variation in racism. That's not the case data wise, but that's an expectation we can hold and then I compare the data based on that expectation. But I do think variation is a big problem in our expectations. So I'm not sure if this exactly addresses but this is another point that I thought might be interesting to make. So sociologists of health or health sociologists sometimes have this notion of fundamental causes. This idea of fundamental cause comes from something like, why is it that in a variety of societies in a variety of time frames, folks at the lower that are defined as lower SES by their societies keep having different and worse health outcomes usually but but low SES is a completely different configuration in different times and in different places. So, you know, one place it's not having a refrigerator or TV is low SES but in another place that's a completely, you know, terrible mod measurement of low SES or, you know, it might be something else. And so what sociologists have tried to think about there is like even with all this change we still have this relationship between class or class status and and health status. So actually, it's, it's, it's that we should not think of those, we should not measure those things in an extremely precise way, or we need to understand and kind of maybe an indexical way or something that is an emergent problem in social in different societies that is not reducible to the measurable operationalizable elements. And so that's sort of maybe a different way of thinking about the way that structures work. But that are, I mean, so this is a little bit of a, like, while even though while I am very assured that you can, you know, measure complex behaviors in in animals in a very precise way, in some ways, maybe a lack of precision or a recognition of the similarities between things that ostensibly look very different in different contexts might tell us a lot more about about outcomes. Jen's question. Yeah, I just wanted to say a few things about complexity. I find time and time again, as a field we tend to shy away from complexity. I'm, you know, the questions we can ask with stats methods and other things, but I do think we're kind of shooting ourselves in the foot in a way. Because if you think about all of other fields that try to interrogate the health of populations, they embrace complexity. It's a feature, not a bug. And our colleagues are much more comfortable with that, which is sort of another reason for collaboration. But then also to the points of medicalization of a lot of these traits. We're relying more and more on biobanks in which things have to be medicalized. We're looking at billing codes, right, which are not there to capture the nuances necessarily of outcomes, but to chase the numbers were sort of collapsing many fee codes down to these categories and sort of maybe that's not the best way even to find the mechanism, right, if you're having this sort of amalgamation of different outcomes into one trait. So I, I wonder about, you know, with these different axes and that the complexity is there a better way for us to move forward that embraces this complexity and acknowledges it in all of our studies, the way that other fields might do so to sort of help us know navigate a path forward that that allows us the sort of breadth of expertise and questions. All right, and I think we're going to let that be the last word. And we'll have lunch. Let's thank the panel.