 So, the second concept is titled Developing Novel Theory and Methods for Understanding the Genetic Architecture of Complex Traits. Alexander Arguello, Program Director in the Division of Genomic Science will give the concept presentation. Thank you. So this concept has come together with Jen Troyer, as well as many helpful discussions with other colleagues here at the institute. So I want to start by reminding everyone of two points from NHGRI's 2020 Strategic Vision. The first one you heard about earlier, which is a guiding principle or value, which is really when we're thinking about research, conducting research, interpreting our research that we keep in mind that both genetic and non-genetic factors contribute to health and disease. The second point is sort of a research goal or objective, which is to determine the genetic architecture of most human diseases and traits. Now recent studies have highlighted the fact that for some diseases and traits, it's actually not clear where the genomic and non-genomic beginning and end, that the line between the genetic and environmental can become blurry, and that for some diseases and traits, determining the genetic architecture of them may not make complete sense or may not be appropriate. So with this in mind, sort of the purpose of this concept is to support novel theory and methods to sort of disentangle both the genetic and non-genetic sources of complex trait variation across individuals, families, and populations. We have several motivating factors going into this. One is that individuals are both passive and active participants in their biological, social, and physical environments. Now by passive, we mean you are born into certain families, born in certain regions of the world, whether they be urban or rural environments, you're born into certain economic strata. And by active, we mean that the choices you make, who you decide to marry and have children with, who you associate with, what substances you ingest, how much physical activity you engage in, those also influences your health and trait outcomes. Now over time, a lot of these family geography exposures and even socioeconomics can become entangled with genetics, so that genetics co-vary with these factors. And these co-variances can actually confound genetic association studies, so it becomes difficult to draw proper causal inferences in what's driving a trait association. So for example, you can have one allele associated with a health outcome that's uncertain whether that allele is operating on your own physiology and biology, increasing your risk for disease, or whether that allele is affecting your behavior, or that then alters your exposure to a toxin, or that allele is found in your sibling or your parent, which then influences your rearing environment and your later outcome. So what we want to encourage is really rigorous theoretical and conceptual frameworks that can help us constrange large-scale data, especially now that we have these biobank-scale data as both at the genetic and phenotypic level, frameworks and theory that can help us interpret that data, guide new data generation, and also formulate robust research questions, because good theory helps you ask good questions and tell you what data you need to answer those questions. But as you heard earlier, multi-disciplinary approaches have a lot of benefits, and good robust theory draws on from experience from many disciplines who are really looking for approaches at the intersection of especially the natural and social sciences, social sciences like sociology and economics, which are used to dealing with the complexities of human behavior and mass, and we believe those types of interdisciplinary approaches will have a more comprehensive account of trait variation in people. So a little bit of logistics. So this is a PAR, meaning there will be no set-aside dollars we'll be drawing from our regular investigator-initiated funding pool. The activities will be R1s and R21s, and these applications would go to our standing study sections from CSR, and one benefit of a PAR though is that it allows us to have special review criteria such as the need for interdisciplinary teams. One thing we're pleased with because we're interested in the diversity approaches, we have many ICs that have expressed potential interest in this concept, underscoring the fact that understanding the genetic architecture of complex traits is sort of a fundamental issue that's relevant to the missions of many institutes across the NIH. And with that, I'll be happy to try to address any of your questions. Nancy, would you like to start us off, please? Yeah, just a couple of questions to help clarify what you're expecting for applicants. Really, you want to stay in the realm of theory and modeling as opposed to data generation, correct? So the idea is to use existing data, but try different kinds of models to look at what it would take, in essence, what kinds of additional data might allow us to discriminate better among alternative models. Yeah, that's what I mean. Yeah, the idea is we have a lot of data now. We're expecting much more data in the future, and we want to get ahead of those data with the proper theory and methods before we get sort of a get lost in it. And I think you had also mentioned in the concept proposal consideration of both demographic histories and also adaptations in the context of natural selection that really have not been part of models, and I think is an important set of additional considerations for modeling going forward. So I think that really resonated as well and would in the community also. So... Yeah, there was a few years ago a lot of controversy over just understanding the genetic architecture of height across populations and how different selection pressures may affect that, and it became a very thorny problem in which we still haven't really solved the way for something as simple as height. So developing better methods for other traits would be important. Yeah, so I think this would have a lot of resonance in the modeling community and would draw a lot of interest. The devil will be in the details and the kinds of people that you get to review, and so that's going to be a... And do you expect to have special study sections or this would go to regular study sections and you'd just be looking at things that target these kinds of problems? Yes, initially it would go to our standing study sections and we just want to see how they do, right, what kind of applications we get and how they're received by reviewers, and then we'll after a few cycles re-evaluate and see if we need more specialized study sections. Howard, I see your hand, but I'm going to invite Laura first. Thank you. So I think that this is really exciting and the whole idea of getting ahead of some of our data is really critical. As I'm reading this, I'm thinking of all of us as kind of a test case with it. And that means, especially if we want to do these really interesting kind of econometric studies, we're going to need a lot of data from all of us about states where people were raised because there's a tremendous amount of information there. And I don't know if that exists yet in all of us. So I just want to think of moving it forward how this could be very interesting, but it's going to mean more data generation from these existing data sets. I know some of these data exist in the UK Biobank already about childhood exposures and where people live, but as I think of these genetic studies, we really need to go back a very far time. Even zip code, postal code gives you a lot of information and cross-referencing that with a lot of existing data helps. And how do we keep confidentiality? So I don't know how all of us is doing that yet. But again, that zip code could be a very interesting thing if it could be matched with the other type of data. So I think combining the genetics now is easy almost. And it's getting these environmental measures that are really going to make this a rich data set and a rich type of study. So I think this is great. OK, Howard and then Olga. OK, great. Thank you for that presentation. I think this is also an exciting concept and I'm enthusiastic for things that I met need. My question's about the mechanism that you plan to have no set aside funding. And so for like novel theories, you may actually get maybe basically different splits in the field that might be controversial. It might have a harder time getting consensus in every study section. And how do you think about that? Do you just want to go for a couple of cycles and see what happens before you decide or how are you thinking about this? Yeah, something we'll have to pay close attention to. Right. Sometimes study section may not be as receptive to novel ideas as others, but we'll have to be careful and see other certain types of applications that are not making it through and then re-evaluate whether we need to convene our own review. Olga, go ahead. So I completely agree. Thank you for the presentation. It's a really important area, absolutely. And I think a challenging one. I completely agree, by the way, with Howard and I can't remember another previous comment on the fact that I would really watch it carefully in terms of having it go to a standing study section and set aside funding, because I do think it might require really, I mean, that's the whole premise, right, that would require different theories and it would require theories that are deeply rooted in applications. And so it's exactly the sort of thing that would be tough to pass through the standard framework. So I would just keep an eye on that. But my other point is I think if you want really new ways of thinking, especially if we're trying to bring in social scientists, it will be incredibly helpful if there are specific data sets that you could point people to. And there are some existing ones, but it's critical that, right, that at least ideally not even zip code, but census tract data, which of course is more highly resolved, is shared under appropriate guidelines. It's non-trivial for these data sets, right, because obviously it's somewhere in these data, but then it requires some very careful cleaning up to make sure that there's no de-anonymization issues and etc. So it's, you know, even though it exists, having gone through in my own personal research, how complicated it is to get it even for consortia where you already are part of them, I would really encourage you to think through it ahead of time and see if there is a data set where you can say, you know, you can use your own data or here is a, let's say, all of us. And here is the procedure that you go through to get access to these sort of data. That's a good point, and we should consider that as we're writing the PAR to include some examples of data sets that may be leveraged. Other questions or comments for Xander? Go ahead. Xander, was there expectation that people would create new models for disease risk prediction as a result of this work? New models? New models to predict disease risk? We, I think there's a lot of work now in terms of models for prediction predicting outcome. I think the hope this is you go a little bit deeper to understand like sort of the mechanisms, all you mechanisms and quotes behind the different factors that are working together, such as selection pressures, different types of selection pressures, or migration pressures, things like that. So it gives you a little bit more granular understanding of how these predictions models could actually even be working, what kind of information they're using. So going beyond just prediction. I might suggest that you use some of the recent work from Molly Przorski's lab, showing that the different mechanisms for how mutations arise actually can both be consistent with a lot of the observations to this point in time, but that and that the more data we accumulate, the less support there actually is for them, the more traditional expectations on mechanism. And the more it seems that other mechanisms are likely to underlie the mutation processes. It's a great example of how pure modeling really showed that much of the early data was really consistent with a couple of different models, but the more data that accumulate, the less established model is supported. And that's the point when we don't have enough data, you think you know what's going on, but the more data you accumulate, the better opportunity you have to discriminate among what can be quite different mechanistic models for things. One of the ideas is can we revisit some of the models that have been around for at least a century in terms of trait variation, now that we have the data, we have the computational resources, the algorithms, can we revisit those models and begin addressing open questions with them? Plus a lot of the sort of bake-offs involve simulations that put imposed genetic models on existing genome data, which were actually shaped by demographic and evolutionary forces that then disadvantage models that try to use those things, because it was just superimposed rather than more organic. And so I do think this is a really important thing for the field to be able to work on going forward. Last call. Can I get a motion to approve the concept? And a second. All in favor? Anyone opposed? Anyone wishing to abstain? All right. Thank you, Zander. Thank you.