 Eric, while you're getting ready, can I answer your question about why the pharmaceutical industry doesn't work on things that kill people? Yeah. Okay. We don't know how. Okay. So that we'd love to, but we don't understand the biology that allows us to do it. So that's why we desperately need to understand that better. I'll expand in a few minutes. All right. So hello from Houston. Thank the organizers for the chance to come. I'm going to begin with the punch line just so we can put it into context. I viewed my tasks and looking at obstacles, and I divided it into three areas. Not in the beginning, but as I started to put things down on a slide. The first were technical and material issues. And I'll be honest with you, I see there are many, many hurdles. So I don't mean to, in any way, minimize the impact of these, but frankly, I see that all of those hurdles we can get over, particularly with the kind of people we've got together in this room, except for maybe me. But they're not walls. They're not going to stop us. It's just going to be a lot of work, and we're going to have to work through many issues. Moving from that, I see there are still our conceptual issues. We've popped up a few times, and there are probably many. One of them is I think we need to have a clear articulation of the goal. Is it to identify targets? And I think we need to be careful and make sure we stay with one or two primary goals. I think too often we try to develop a resource like this that's a one-size-fits-all, we end up maybe not achieving our primary goal because we try to make everybody happy. And so I'd like to see us during the coming days, and I'm sure months ahead, make sure we're very clear about the goal. The other, and Aravinda alluded to this, I'll say more about it later, is organizational issues. I think there are a lot of cultural and organizational issues that are really going to be the most difficult, and they're probably uncomfortable to talk about. And so particularly this group, I think it's important we get them out on the table and we look forward. So what are the technical and material issues? Is there a pointer? No? That's okay. So what I've done is the next couple slides, I've listed just a few of them. In my opinion, there are more hurdles. They're not game stoppers. First are there the available samples? And I think the answer is yes. Thank you, Lisa. So what I've done here, and this is published in an editorial in genetic epidemiology, is I put together a large number of cohort studies, and this is aged down here beginning with basically birth and children going all the way up to octogenarians and centenarians. And there are a lot of samples that are available. And these are very well phenotype samples. So I think as we move forward, we need to look at ethnic diversity, cultural diversity, age, if you notice, not a lot of teenagers in the 20s. And so there are holes in this. There's a lot of kids, and there's a lot of older people where we tend to see early onset disease and chronic disease. But I think once we better articulate the question, we probably ought to ask what do we want this sample set to look like and make sure we fill holes. And I think this even gets better as we think about international collaborations. The UK was brought up and also in China. The other is data sharing, that we tend to think about these large population bases as not wanting to share data. And I think that's not the case. We can complain a lot about dbGaP, but dbGaP along with the GWAS era, I think has really set a wheel in motion, a culture of data sharing. And so as we move forward, there are words in the documents that were circulated like streamlining or creating a new venue. And I think that's fine. Do we need to streamline this? Absolutely, we have to streamline this actually to achieve our goals. But the key that I wanted to get across is the large population data sets or the large population studies, they've crossed the bridge about data sharing. I don't think that should be an issue. The other in Gonzalo, I think brought this up earlier, is the complexity of the data. And this slide, this happens to be the pipeline in the genome center at Baylor and Houston. And there are similar pipelines at the Broad and I know in Ann Arbor also. But I think we have the structure basically to bring BAM files together from diverse resources and basically homogenize and get the map to a central or a common map. And then also to call variance in a common way and then distribute that data back to this it, whatever this is, it's gonna be a commons or dbGaP. Again, I think there are many people working on this very difficult problem. So again, this is not I think a game stopper. The next issue, the next set of issues that are conceptual issues. The first is I still think that we're struggling as a field to think about the impact of rare variants in public health. And people are quite comfortable about rare variants in Mendelian disease. I think people still think that rare variants have these very large effect, that I forgot who said it, that make a foot come out of the top of your head. But indeed these rare variants have effects in the continuous distribution. So you'll see enrichment of rare variants in the tails, that's clear. You'll see clear Mendelian disease causing variants in the tails of the distribution. But you'll also see them throughout the distribution for many reasons. One is variable penetrance. Second is the effect, where did they start in terms of their background environment or their background genotype? So they're not always in the tail of the distribution. So I think many of us need to, I think, think harder about the role of rare variants in health and disease. The other is we need to recognize successes. I think in general, the last, I don't know, five years or so, I think we as a field have not done enough about marketing and trumpeting our successes. We've had very many successes, and this is one that's already been brought up. I think it was Francis, if I recall, brought up that worked with Jonathan Cohen and Helen Hobbs in the Eric study where we've identified basically loss of function mutations in PCSK9 that lower cholesterol. So these are people that have the variant or have a variant. This is the people who do not have the variant. And you see two things. One, you see it lowers LDL cholesterol. And going back to the previous conversation with David, is it also lowers events? So this is basically the percent with coronary artery disease or coronary heart disease. So not only lowers risk factor levels, it lowers events. I think far more importantly, frankly, is the ability to follow these people and look for adverse effects. So we're people that have these variants. We bring them back into the clinic and we do a much better job in phenotyping and look for adverse events and adverse effects. And if you follow the literature, for example, recently on statins and age-related cognitive decline. So unfortunately statins increase the rate of age-related cognitive decline. I can also tell you PCSK9 carriers have advanced rates of age-related cognitive decline, just as an example. Most importantly is organizational issues. Francis isn't here, but I stole this out of a nature perspective. I think there are numerous silos that we're gonna have to begin to dismantle to make this it a reality. The silos exist at the NIH in terms of funding agencies and NICs within the NIH. They exist within each of our institutions. Believe it or not, they exist within very well-functioning consortia. So we're gonna have to remove the silo effects. And I think one of the disadvantages we're going to have and challenges we're going to have, my experience is when money gets tight, those silos get higher and they get thicker walls. And so we're gonna have to figure out creative ways to continuously erode those silos. The next point is we're gonna just a reminder is restating that we're gonna have to have clear goals and expectations. And I use GWAS as an example. It's almost become kind of cool the last couple of years to criticize GWAS. And I think part of that is because it's a moving target, what was it supposed to do? And because it didn't meet somebody's ideal, it didn't explain all of the heritability or it didn't create a drug tomorrow afternoon. There's a lot of press that quote, GWAS didn't work. I read that over and over again and I think that's total nonsense. As a person who spent a lot of their life in complex disease genetics, I can tell you this sort of bubble karyotype that Terry's group has created. It was extremely sparsely populated before GWAS. And so as a result of GWAS, we have a number of loci and in some case a number of genes of which we can actually examine very carefully for rare variants that are contributing to disease. So I think again it's important for us to just articulate the goals and expectations and not deviate from them. I think the other and I'm gonna use something close to home for me just as an example is that the consortia themselves are changing. And I actually think many of you are worried about the culture of data sharing probably more than you need to because the consortia realize we're probably going to need to share data more than we've had. So here's an example from the charge consortium. And then I won't go through the details of this because it's not important for this meeting. It's basically a lot of NHLBI large cohort studies that have done GWAS. They've contributed the data to a central resource where meta-analyses have done. This has been very successful for a very large number of phenotypes. But we realize as we begin to analyze not just exome data but whole genome sequence data that that model is making our scientific endeavor fall short. And so we're moving then from a meta-analysis model really in creating the scientific commons where the samples are sent to a centralized lab. In this case with Richard at Baylor and the sequence is created and it goes into a scientific commons where it's Q-seed. We've had to work through the data use agreements and also all the IRBs. Again, they were hurdles but they were not walls that we could not get over. The only comment that I'll make because it hasn't really been talked about is the analysis model. We've found as we've pushed this forward in the case of whole genome sequence that actually a combination of a central analysis model and a distributed analysis model works quite well. So for those institutions or collaborators that do not have the local infrastructure we can provide centralized analyses and get them results according to a pre-specified plan. And for those institutions that either have a very specific analysis that they want to do that's very specialized we can actually distribute the data back out to them. So this combination of centralized analyses and distributed analyses I think works quite well. The other is not to forget the important cultural truism of making sure we have good young motivated investigators. Many times if the project is too centralized, young people tend to think that it's an imprenetrable barrier. So we need to make sure whatever we create that we invite to make sure we foster young investigators who are going to make the most use of these data and also translate those into publications. And here's just an example from that. Here's whole genome sequence data happens to be CETP. I picked it just as an example with HDL cholesterol and you can actually then use this kind of model. Then this happens to be from a centralized analysis with Eric Framingham and CHS, the analysis using whole genome sequence data. So I think this is an exciting direction that we're all heading and we have documented to ourself that some sort of a commons is necessary. You cannot do these types of analyses using a meta-analysis approach. You need to be able to bring the data together. So just in closing, I'm an optimist and I would leave us with two things. I'd say we've met the enemy and they're us. And I really think we need to work on these cultural issues much more. If we're gonna have the scientific commons or we're gonna have a centralized data set, we're gonna have to work together to lower those barriers. And then I think the other, and HGR is very good at this, is setting very large goals and sticking to those goals and breaking those goals down into a series of small steps. I think we've got the right group here together to make progress. And I'll stop there and entertain questions and comments. Lisa, do you want me to leave this up here? Irvinda? I'm gonna ask a question just more to the science in the future. I mean, you're much more aware of the details, which is, what is the status of cohorts, of getting cohorts and getting new people into cohorts or new kinds of cohorts today? I mean, much of what we're meaning really benefited from are the cohorts that have existed, funded in the, quite in the past. And I'm not saying it obviates any other thing we ought to discuss, because there will be new samples collected, nevertheless. But the cohorts have a very particular kind of place in this kind of new epidemiology, so. Right. Well, I mean, it's not a secret that cohort studies are expensive. And it's, and also that money's tight, so. But on the other hand, I think the leadership has examined the existing portfolio of cohorts and asked where there are shortages. And one of the probably the obvious one that NHLBI and I think NIDDK, and please correct me if I'm wrong in that, have come together and formed a large Hispanic study known as SOL, the study of Latinos. Because that's an example of an area where it was almost an embarrassment. We have the fastest growing sector of our population. I live in a state where we are a minority. They are the majority in that particular case. And not only that is they have a disproportionate burden of disease. And so there are new cohorts being founded in that study of Latinos as one example. And I don't know if there's anybody here from the Children's Institute. There's the National Children's Study, which has limped, I would say. But I'll try to be positive. There's a lot of effort to kind of re-initiate or recharge that under a more simplified model. So it spends less and collects more individuals. The only comment that I will make though before I stop is on the recontact issue, I think it's an important one. Having worked in many, many cohort studies over the years, I would say, and collaborated with geneticists with those cohort studies, 99 times out of 100 it works extremely well. The one time out of 100 it doesn't is some cowboy tries to contact the participant themselves. And it's important to first not to do that and to set up mechanisms so that doesn't happen. That doesn't mean you can't recontact them. We have permission to recontact them, but there's procedures, appropriate procedures to go through to recontact them. So I think whoever said it, we probably should have a separate meeting about recontact, but that doesn't mean we can't recontact. We can. It's just we have to go through the appropriate steps to do it. Yeah, I'm interested in this discussion about cohorts because in Europe, there's a real shift towards just using your health care system as your recruitment system, as your blood tracking system, as your phenotype, as your first line phenotyping, and this obviously requires electronic health care records and coordination across that. And although that's not universal, there are projects obviously, I mean most obviously in Iceland, but also in Finland, in Scotland, and in Denmark, all set up in a very similar way. And this seems like the kind of, this seems like the end point. And so why not go to the end point? Well, if you think we're going to come here and change health care in the US in a day, it's not going to happen. No, sorry, let me get this straight. I guess using HMOs, Mayo Clinic or whatever or the Veterans Association, why not use the Veterans, is that the right term? If you think we're going to have the Veterans Association in a day. So there are a lot of issues in access to these types of data in the US. But again, let's be positive. The US is moving towards a more electronic medical record that has a semi-standard format. I think the difficulty in the US that electronic medical record tends to be for billing purposes, not really for diagnostics purposes. I also think is this discussion moves forward, it's Eric probably can correct me representing NIH. My understanding this ID is not a US yet. This will be a collaborative or national effort. And my hope is that we'll have studies from the US, including some of these cohorts and many others. Studies from the UK, studies from Iceland, Holland, China, etc. So I think we'll have a mixture of formats. Yeah, so Eric, I think your mixed model is extremely attractive. Because instead of having a fight over should it be centralized or should it be distributed, identify which things make the most sense to study centralized and which things make the most sense to be distributed. Certainly the phenotypes being distributed is extremely important because it's going to be very hard to get rich phenotypes centralized. On the other hand, DNA analysis is much more easy to do centralized. So I think that despite the disparities in the healthcare system, to see the cohort groups that have the richest phenotype data coming to this kind of solution is reason for optimism. Yeah, I would agree. I was going to comment that for example, Kaiser of Northern California has had a GWAS study done on 110,000 and is another 100,000 or so waiting to be done. And so I think the idea of using medical records in HMOs is really an important piece of this complex future. Of course, the good aspect of it is that the phenotype thing's already done and it's available in an electronic medical record. The bad side is that the phenotype's already been done and it's in an electronic medical record. So you don't get to ask the questions that are not being necessarily done as part of routine clinical care. I'll make one comment and then I'll turn it over to Debbie. We have an enormous sector of this population that's uninsured. So to think that every American has access to Kaiser is just not the case. So I was just going to bring up to answer you in question is that there is a large scale project in the U.S. called Emerge that actually is looking at genetic association. So there is a mixed model already and I think it's really important to consider this rich model as Bob has brought up with Kaiser. There are many HMOs in the United States that are exploring these issues and doing GWAS and showing it can productively be used to find similar things that can be found in the cohorts. And I think that's true for every level of genetic data, it's just a matter of learning how to use it. And obviously other countries have been more facile using it than we have, but we are looking at it. The other thing HMOs may solve is this gap of age. If you looked at that first slide, the number of individuals studied between around seven years old and 20 years old is very, very small. So things that are occurring in young individuals, we're missing if we were to rely on cohort studies. If I could just make one other comment about culture. And that is that David made the point that many people in pharmaceutical industry don't know biology. But I think that many biologists don't know any computational genomics either. And in fact, this reminds me very much of what the situation was like when molecular biology first started to move out of basic science labs and start to be applied. I remember distinctly a visiting specialist expert came to our lab when I was a postdoc to help us learn how to make packaging extracts for phage and how to make ECHO R1 from bacteria. And I think there's an enormous educational gap in that perhaps one of the things that can come out of this meeting is a very, very strong push to increase the size of that pipeline. I know there is a pipeline, but it's much, much too narrow. Well, I think one of the things everybody would agree, this is going to be an interdisciplinary effort. This isn't going to be solved by either computational biologists, by genomicists, by epidemiologists. It's going to require an interdisciplinary effort. There was one question over here.