 Hi. Thanks for the opportunity to share some work that's ongoing in cystic fibrosis, which is a, we're able to go to the extremes of phenotype patients at UNC and pick the extremes of the extremes, stratify by gender, and key genotypes of risk and non-risk in the thing, and got Debbie Nickerson's group to RS and G sequence of 370 patients across this 300KB region. And interestingly enough, we actually added, what is that, 116 SNPs that were of a sufficient frequency to cause imputation. By adding these additional SNPs in, we were able to get a better imputation in actually the association P-value in this set of patients homozygous for Delta F actually got stronger. I will say that, I will say that these additional SNPs, novel SNPs, in fact, overlay and are in or near these open chroma in the regulatory domains, and I can say that in pilot preliminary work where they inherit, we've at least one of these, we've already identified at least one strong enhancer with an effective 20-fold effect on expression of at least one of the genes in this region. And so the home run would obviously be if one of these SNPs modified the effect of that enhancer, and we're going to work on that, we're obviously going to follow this whole region up in a systematic fashion and try to identify the regulatory elements. These interactions, as Steve was talking about last night, by using a chromosome confirmation capture, try to determine the function of these regulatory elements. And then we downstream have human airway cells, which we'll be able to do a variety of experiments to try to understand the mechanism. A second finding from the original 3,500 patients was a very strong linkage peak on chromosome 20 in about 500 twin and Sib pairs from Hopkins. You can see that a lot score is almost five, and if you actually put in a BMI as a covariate, the lot score is greater than five. As you can see here, this is again over a non-coding region. There are some interesting genes in the region which do relate to lung function, structure, and inflammation. We were able to go under the linkage peak and look in the association patients from UNC and Toronto, and for a region-wide analysis, we were able to replicate this linkage peak in that population. I will say as a point of provocation, we're currently doing a study with Ed Silverman looking at COPD patients and trying to look at commonality of SNPs and genes that we find in CF and COPD, and our second ranked SNP happens to lie in this region. So we think that's a very provocative observation. Gary Cutting is leading the charge on this, and we'll be sequencing a large number of patients across here. The concept is that probably can't explain this by common variants. There are probably rare variants in here, so it may be a combination of common and rare variants that are calling the tune. We have participated in the exome sequence. This is work that's largely derived from Mary Emans and Mike Bamshed in Seattle, where we tried to look at the age of onset of pseudomonas infection. The early age of onset of pseudomonas is a bad prognostic sign, and we had some of our patients in this exome. And as you can see here, Mary did the traditional burden analysis in polymorphisms that were less than 0.125, found one, the clear outlier on the QQ plot in 91 exomes went to 696 additional patients and sang or sequenced this gene, D and DCTN4. As you can see here, if you have a missense mutation in the DCTN4, you have much earlier onset of pseudomonas than if you do not have. So this was a replicated finding, and this work is actually impressed in nature genetics. I will point out that we have well phenotype patients with these other disorders, CF-related diabetes, 25 to 30 percent in adults. Scott has already shown that the top ranked SNP associated with the CF-related diabetes is TCS7L2, that old familiar one that causes type 2 diabetes. There is a data forthcoming now from the GWAS analysis, and he has identified at least two other type 2 diabetes genes associated with this, as well as a novel, a PsiU transporter. It turns out that PsiU transporters are clearly associated with myconium ileus, as published just a few months ago in Nature Genetics. Patients who are born with myconium ileus, and it turns out that there is the same PsiU transporter affecting myconium ileus and diabetes in CF patients, suggesting the possibility of a pleotropy, which we think is quite interesting. Five percent of CF patients develop portal hypertension. We have already shown from Canada gene studies that if you carry one copy of the alpha-antiprotease allele, you have a four to eight full increased risk of developing liver disease. If you have CF, and the effect size is bigger than the effect of the zeolial in patients who have alpha protease deficiency. So basically, we think that there are already some neat things going on. CF is now becoming a complex genetic disease, even though it's driven by monogenic recessive disorder. As I said, we have an additional 3,500 patients who are undergoing GWAS. We have GWAS data in these other heritable traits. And so we think this population is a fertile ground for discovery. And of course, acknowledgements all the people in the consortium support for the USCF Foundation, which has really funded a lot of the GWAS stuff, the help of NIH, NIHR, ICNG stuff, et cetera, et cetera. Thank you. Thank you, Mike. I wonder, perhaps, Nancy or Peter could comment, how would the strategy like this work in the large courts we are discussing with exome data, where the main effect size gene is probably lower penetrance than CF? And the total numbers, the n's will be smaller, though the overall numbers are again large. And can we do GWAS, will we be okay with GWAS and exome data, or do we need whole genome sequence data? Would it work? I mean, in terms of GWAS, a lot of the action has been in SNPs that aren't in genes. So I think there are big advantages of doing whole genome. I don't know whether we're going to discuss it specifically. My view on sequencing of large, if we're going to do a large project, sequencing of large cohorts, I think the arguments for doing whole genome sequencing are very, very strong over exomes. I mean, not specifically for this reason, but it's one of them. We will be doing everything with whole genome sequencing at some point in the relatively near future, I think. And if we think about the value of such a resource, if we were to do a large project just on exomes, then we'll be very frustrated in two or three years' time, because there'll be all sorts of stuff that we don't have in the data. So I really think if we're thinking of a large project, it makes so much sense to think in terms of genome sequencing rather than just exome sequencing. Can I, sure, I mean, I'm going to shift back to a different point that I think might raise very nicely, and that is the ability to look at related phenotypes and ask the question, because if pathways are real, they probably aren't working in isolation only in this disease and in this particular setting, and I think it's a very nice example of you're talking about going from CF to COPD, yes, they're very different configurations, but they most likely are going to share certain things. And so in thinking about the larger structure or the larger study, I'd like to just put on the table that I think that that would be a very useful and important criteria to consider that would be these sort of overlapping outcomes where not only are you looking at one particular outcome, but something that may be related that allows you to do this. And we know that in the history of GWAS, I mean, we've been lumpers to find things, whether it's within cancers or BMI or inflammatory diseases, but that by lumping, we're able to have the numbers to be able to really achieve a level of discovery that pinpoints new pathways or new places to go after, and I think, and then, you know, all the mapping and exciting stuff that goes afterwards, but I would just like to put that on the table and emphasize that because I think that was really an important point. Nancy. I think these findings also raise some interesting points again around, you know, back to efficacy and even adverse events. So here you have TCF7L2 coming up in the context of cystic fibrosis associated diabetes. When hypertension, for example, is an adverse event in treatment with VEGF inhibitors, why might that be another way of identifying hypertension genetic risk factors? Similarly, you know, with the efficacy, I think part of what will be the genetic, will be the genetic risk factors that we find that affect efficacy will just be about the genetic basis, the underlying genetic basis for the phenotype. We basically sort of have to fix what's broken in order to have good outcomes. And so these are all these cohorts that are well phenotyped when there's outcome data. We just have that much more leverage to getting at some of the same underlying questions that we think we want to get at with just the disease phenotypes. So I think it's really critical to have the outcomes phenotypes for in these cohort studies. Question. On the TCF7L2, you said it was the strongest variant for diabetes in CF patients, but it's the strongest variant for diabetes, at least identified by GWAS in non-CF patients. So is the odds ratio or the effect much larger? The effect size is four to six times greater in CF patients than it is in general population. So it's like an odds ratio of like three? Yes. It's four, it's actually four to six. It's a little complicated because CF patients get steroids to treat their lung disease. And if you control for that as an effect that induces diabetes and take that out, the effect is actually close to five, some of them are between five and six. So that all converts to around four. Cool. Moving along to what Nancy just said, so has anyone then looked at steroid-induced diabetes in non-CF patients in relation to TCF7L2? I know there's been looked at in relation to metformin, but I haven't heard of it being looked at in relation to steroid-induced diabetes. Oh, I'm sure it has. Scott Blackman at Hopkins is really leading this charge and I can't answer, I'm sorry Terry, I can't answer your question. Yes, and this is in CF patients who obviously do not have obesity. Yes. So despite the fact that they're not obese, in fact, if anything, they're undernourished. They still have the development of diabetes associated with this polymorphous. Out of comments, questions. So one of the reasons to have a classic Mendelian-ish disease on the agenda was to address how can we address Mendelians in a large sample, even in a million people you're not going to have many CF patients. So it may be that, you know, obviously no one study is going to answer every question. But is there some value in looking at phenotypes of heterozygocareers, it sounds like there is, as well as, you know, formfruits, et cetera. So could you comment on that? The specific question you want me to ask is... So how would we work in addressing Mendelian diseases in a essentially population-based or unselected cohort? I think that's the question. Well, I think it would be extraordinary to take all of the families that are well phenotyped and do sequencing so you have affected and you have parents. And for CF patients, and you can control the CFTR genotype, and I think it would probably provide extraordinarily useful information as we're dissecting through all the additional genetic variation that we're going to identify. I mean, the chromosome 20 region is a great example that's going to need to be sequenced. So the... Well, another potential application would be to reduce the multiple testing burden. So you can imagine that if you were to look at, say, kind of hepatic or pancreatic phenotypes of interest, you could argue that because of the mutations in CFTR, or the side effects in some CF patients of those things, that that would be a logical thing to look at when you're aggregating, you know, across pathways or across side effects or organ system side effects. So one thing about Mendelian diseases is not all made equally. So we have Mendelian diseases actually give you a complex phenotype. Actually, the lung phenotype in CF, you can call it complex disease, almost. So that's where the rare Mendelian disease can link to the common complex disease. We can even make some major theme to cause common complex disease. For instance, the CF, the overlapping with the COPD, maybe it's about lung injury. It's all some kind of lung injury. And after lung injury, there's fibrosis process going on. So maybe it's fibrosis. So in that way, if we can start to link or mix the data from rare Mendelian disease, which has complex phenotype with the common complex disease, we may actually... We can fit each other and move the field a lot faster than simply moving, you know, separate the two field and now to combine them together. Eric. Maybe Nancy, I'll put you on the spot. I'm kind of amazed that we act like there's Mendelian diseases over here and sort of Peter Vischer polygenic disease over here and there's nothing in the middle. And I'm wondering if we need to be reinitiating old discussions about oligogenic models of inheritance and even two genes. There's very few of us even doing two gene models. And we need to come back to that if we really had a very large sample size where we could seriously address these questions. Maybe we could think about two and three sort of oligogenic models of inheritance and disease susceptibility. Before you answer, let me just add, you know, there are complex diseases like autism where there's also evidence that common genetic modifiers may play a role. So in the area of more development, there's a lot of evidence that this could be the case. I mean, we used to think that oligogenic was... I mean, recently had been thinking that oligogenic modeling was, you know, sort of a way of fooling ourselves about things being simpler than they really were. But we do recognize gradations. I mean, even among our complex phenotypes, you know, there's Hirschfrans and there's Crohn's disease and there's Type 2 diabetes. I mean, if you consider the amount of genotyping that had to be done to discover a certain number of genetic risk factors, it's very different for those phenotypes. Approximately the same number of genetic risk factors have been established for Crohn's disease as Type 2 diabetes, you know, highly significantly and reproducibly associated with disease. But it's taken, you know, about an order of magnitude more genotyping probably to do that for Type 2 diabetes than Crohn's disease, at least, I think. And that's a function and part of the genetic architecture that we are starting to get an understanding of. And, yeah, you could think about things more in the Hirschsprung range if you wanted to get, you know, maybe take a shot at a complete understanding of something that was intermediate between complex and more Mendelian. I mean, I completely agree with Nancy about Type 2 and Crohn's. I just, I see that as both having complex genetic architecture, one of them just has more larger effects, well, they're not large, but less, more slightly larger effect sizes than the other one. Are there any examples of two gene diseases? Three. You asked. Or three gene diseases or five gene diseases? Five, yeah. Let's take a look, too. There must be, Adam. There must be. Well, tell us what they are. Andy is close. Second generation. Two. Gene. I mean, both are gene, but that's the challenge. I want to get at what Eric was trying to get at when you were asking. Do you have specific examples in mind? No, I know that I'm wondering how many people are not doing it. We tend to look at them in the alien. We tend to look at them as polygenic. And we're not fitting, or we haven't been fitting. In the old segregation analysis days, we did fit two gene models, but we've gotten away from that. And I'm wondering if we stop fitting sort of one locus at a time, we go back to, and maybe it's a way of better integrating families, should we go back to integrating sort of countable numbers of individual genes, whether it's two, three, four, I don't know. But we tend to look at one locus at a time. And I wonder if it's a way of adding information by looking at two, three, four. I think there's an extraordinarily good example for these ciliopathy, filled with ciliopathies. I started off working in motile ciliopathies, because cilia work in the airway. It's mucociliary clearance. It turns out that they're a sensory cilia on almost every cell in the body. And you get mutations in genes there. And you get there at least eight different types of Barty-Biedel syndrome, and renal cystic disease, et cetera, et cetera. And Fredham Hildebrand up at Ann Arbor has got this chart where he talks about, in result, clinical phenotype, and how many of the different genes in the pathways contribute to these net things. And so I think the ciliopathy are probably a good example of the kind of things that lead to similar but different clinical phenotypes. I have a suggestion. So Gail has inspired me to think about quantitative trait, which is great significance to all of us who go to primary care offices, which is LDL cholesterol. Obviously, one of the largest treated conditions. And there are a limited number of Mendelian alleles that impact LDL. In fact, some of the data that we're putting together in the exome sequencing project for LDLR, that is being confirmed. There's not a huge number of new genes. ApoB, LDLR are the drivers. And yet also in the more complex, less extremes of LDLR, we know from the global lipids that there's what, 35 GWAS SNPs. So that may be a nice quantitative trait. And others like that to look at this question of where the Mendelian begins and the more complex interactions of genes begin as well. Lin, you had to come. I was just going to mention that there are at least a few cases of truly di-genic diseases. Retinitis pigmentosa has a form that is di-genic. I think Bardae beetle also. So one could set up an experiment, basically, taking a whole cohort of, say, RP cases, where there are a couple of dozen genes at least, and ask the question, how hard would it be to find those cases that are actually di-genic? So it would be, I think, an interesting experiment that it would at least begin to address that question and that issue. Thank you. I think we are taking a break now, a long break. We have till 10.15. Thank you.