 Well, first of all, thank you for the opportunity to be here today and give a perspective of target validation using the vast amount of genetic data that's available in the context of pharmaceutical target selection and ultimately drug development. We've been partnering with Abroad and David Alshuler now for in excess of about three years. And it's really been in the last six months, I would say, that we begin to realize how disparate our conversations were initially with regard to expectations and how we might be able to help each other. And since we finally posed what are the fundamental questions that we're trying to get at, and we shared that kind of information, which seems so obvious at the time, we've really made some vast improvements in the utility that we're seeing from access to some of these databases. And so I'll take you through essentially how we go about selecting targets, what we consider to be a valid target, and how we're using human genetics. About two years ago, we had a pretty substantial reorganization. I'm in the cardiometabolic space, and it came from the powers that be that we were no longer going to pursue targets that were not founded in human genetics, at which point all the biologists and physiologists, including myself, threw up our hands and said, now what do we do? But it is absolutely our mission now to go after those targets that have very strong human genetic evidence to support them, and it's a reflection of the fact that we fail in phase two 50% of the time for lack of efficacy. So we can come up with great chemical matter, we can design great studies, but we're not very good at picking targets. In fact, it's a coin toss, essentially. So this is what I was asked to touch on today. So how could a database be used for drug target validation? What are the types of data that are required? How important is it for us to be able to re-contact these participants? And would you use this database or these studies differently depending upon what you were trying to accomplish? So this came out of last week's meeting in the Information Commons group, where we were talking about the different definitions of target validation. And I really work in the space of target identification at the bottom there. And it really is the generation of scientific evidence that a particular target is involved in a disease process, and we actually raise it up a level higher than that. We start at pathways. There's so many different ways that you can intervene in a pathway, so many different targets to choose from. First thing we just want to understand is, is this pathway going to have an effect on disease progression or reversal? Once we've kind of posed that question, then we start, or traditionally we would then start looking for animal models that had been developed that had tested that particular pathway to give us confidence that we had a good idea and it would translate to human efficacy. And human genetics, frankly, was an afterthought. We relied almost entirely on the animal models at that particular point. And I won't go through all the other definitions, you can look at them yourselves. But we came to work with the Broad through this realization that we needed to have a better understanding of the role that the human genetics plays in disease susceptibility and how we could use that information to select targets with a higher probability of success in phase two. We have a precision medicine group in Pfizer that is composed of a number of geneticists, quite a number of geneticists. And the question is, well, why don't we just do it all in-house? And one of the fundamental problems that we ran up against is, well, we just don't have access to the data. Or it's difficult to get the data. Or how do you combine all these different data sets so it ultimately becomes much easier for us to partner externally and at the time thinking that, well, if we would partner with an institution like the Broad, all the data sets would become quickly available. We could check the boxes, find the targets, and away we go. That's actually a little bit naive on our part. But the goals of the collaboration are stated up there. And it really is to validate human relevance of the targets in question. And then to apply novel methods to bring or to bridge from patients in the human population to biology that we can actually go after. Because the other problem that we run up against is when we see these GWAS studies that are published, they show or report all of these susceptibility variants in diabetes and cardiovascular disease, many of which are just not drugable, or they are associated with progression of disease. But from our standpoint, we have to be able to go in and show reversal of disease. So it has to have a very big functional effect. And then ideally, we could run a phase three study, a multiple year study, that shows not only in the short term can you reverse disease, but over the long term, you could slow the onset and progression of disease. And that's a big challenge for us. Of course, we all want to stop the onset and progression of type two diabetes, but when we run a clinical trial, we have to go first into a diabetic population and show that we can achieve a 1% improvement in A1C in 12 weeks or less. So the target has to have a lot of horsepower in the first place before we can even think about modifying disease onset and progression. So one of the terms or terms that we've come to use with David and the group there is this unbiased versus biased target validation. And the two kind of conversations that we're having are, can you find me a new target backed by human genetics? And so this would be going to the recent publications, seeing new papers that come out. GRB 14 is a recent example that came out last year. It's been publicized again as being involved in the onset of diabetes or susceptibility. So the question is at that point when you see a variant associated with the phenotype that you care about, do you go after that as a new target? It's a novel target, it hasn't been tried before. It would fit the criteria of being novel and potentially disease modifying. But we still don't understand whether it's going to have the horsepower, if you will, to really be a viable therapeutic strategy. The other conversation that we have is, we have a whole list of target hypotheses that may in fact be independent of any of genetic evidence. But we would want to know, is this target that I've already started to work on, is it supported by human genetic evidence as well? Or is this pathway have additional clues from the GWAS studies, et cetera, that would give me confidence that I've made a good decision to go down this particular track? And so what we see here is kind of the traditional GWAS. We have a patient phenotype type 2 diabetes, for instance. We identify a variant, and now we're looking for a target. And we refer to this as kind of the forward genetics approach. This is not the approach that we traditionally take. Absolutely not the approach we traditionally take. We do the reverse genetics. We have a target based on an animal model, associated with a pathway that we think is important. And we start looking for genetic evidence to support our good idea. And in the absence of that genetic evidence, we still chase that target because we think it's a good idea. And phase 2 success would tell you that's not a viable path forward. So back to the GRB14 story, it's interesting about this ability to combine all these large data sets is when the GRB paper was first published last summer, we weren't enthusiastic about it, but it had a very small effect on disease susceptibility. It was in the Southeast Asian population. We didn't know the relevance of that population to the broader population that we would be developing a drug for here in North America. But then following that particular publication last month comes out another study where they combined a number of different data sets. And now you're looking at, I believe this was north of 90,000 individuals in 50-some studies that had been looked at where they had additional phenotypic data and what they were able to demonstrate that in a population that we really care about in the cardiometabolic space with an elevated BMI, you do start to see a big change or separation in the susceptibility associated with this particular variant. So it was this kind of information that takes us from being modestly interested in pursuing this particular target to, well, this might be very appealing, because in the population that we're aiming to treat, it appears to have an even bigger effect. So should we pull the trigger and go after it? And when we combine these large data sets with that phenotypic data, and back to the earlier comment, are we interested in a genetics database or a metabolomics database? The answer is absolutely yes, because these studies that we see that come across our desks in the form of a nature paper or what have you, it shows susceptibility to disease, but it doesn't tell you much else about what that particular variant could be doing to the system. But when we start to combine some of these larger data sets or combine multiple data sets where there's additional phenotypic data, we start to get a very rich picture for how a particular perturbation can influence the entire cardiometabolic profile. And here's just an example of work that we did with the Broad where we went through and we looked at a number of different studies where there was a fair amount of phenotypic data, and that we see as we go from the heterozygic to the homozygic allele, as we move across things that we care about, such as blood pressure, LDL, cholesterol, triglyceride, HDL, fasting glucose. Glucose after a glucose challenge at the 120-minute mark, all of a sudden you can start to put together a compelling story that this particular variant plays a critical role in all things that we care about in the cardiometabolic space. But in the absence of having that database to go to, to look at these specific studies, to look at the phenotypic data, we're left with increased susceptibility to type 2 diabetes should you choose this target or not, with probably very little information about the functional effects. This kind of information for us starts to tell us a lot about potential functional effects of the variant that we can then make a more educated decision as to whether or not we want to pursue a particular target like this. So one of the things that we did with the Broad was we took this approach of we have a portfolio of roughly 20 programs in the cardiometabolic space. They range from early target discovery all the way to phase 2. Did we make good decisions? Was this a good portfolio in terms of is it founded or does it have supporting genetic evidence behind it? So we gave a list to the Broad. Here's all the targets that we care about. We have active programs in many of them. You see that they intersect with fatty acid metabolism, insulin signaling, metabolism of triglycerides, et cetera. Much of this based on what the competition is doing, what's been published in preclinical animal models, knockout models, different cell systems, et cetera. And then we ask the question how much of this, how many of these pathways or targets are supported by human genetics? Which thought was going to be a very easy question. Throw it over the wall and we would get quickly a spreadsheet that would come back and say yes diabetes, no diabetes, fasting, glucose, insulin. And it frankly wasn't the case. And so from a pharmaceutical perspective, we're wildly enthusiastic, I would say, about this idea of having these kind of lookup tables or access to these large databases where you could take your portfolio, ask the question is it supported by human genetics and the indication that I care about, as well as what are other things that should be safety concerns? Because now we're starting to go after a lot of targets that intersect metabolism and inflammation. And all of a sudden you have a target that you think is going to be very good for improving insulin sensitivity in the adipose and then you later come to find out that it's associated with an elevated risk of Crohn's disease or IBD or RA. And how do you screen for that using animal models? You simply can't. So I guess I would close with, from our perspective at Pfizer and we've talked to a number of other pharmaceutical companies that have been interested in going into pre-competitive consortia with us in academic groups or the NIH, do we want a database where we can compile all this information and have access to it? Absolutely. Do we want phenotypic data in there? For us it feels like a must because it gives us so much more information about the functional consequences of these variants or these signals that, you know, that's how we prioritize our portfolio. But you know, we'll take a variant to a pathway to a phenotype that we care about and then make the decision, do we think it's going to be disease modifying or not? So thank you. Please. So this is the path that you and Pfizer are taking with really focusing on human genetic validation. Is that true of your sister companies? Or is Pfizer at one end of the spectrum? Or how does that look? So we are having active conversations with a couple of other pharma partners that have either done this on their own or have done this with other academic collaborators and see the utility in it. They share the same frustration that we do. We simply don't understand how to access the data and even what data are out there. So I would say there's three or four companies that we're in active discussion with that would like to be involved in something like this. It's not clear how we progress forward. However. And just to follow up, that's three or four out of how many big companies? 25. But that's not to say that 22 have said no. These are companies that we have pre-competitive type of relationships with. So I could tell you that really in no instance have we floated the idea that would you be interested in going into a pre-competitive consortia looking at expanding the use of genetic data to pick better targets? The answer has always been, that's a good idea. We want to do that. But how do we do it? It really gets to the logistics. David. I just think the only other comment I would add to that, I do think there's a lot of interest in this in pharmaceutical companies. I think we have to be clear. It's not GWAS, per se. Like it's human genetics. So I mean GWAS is one kind of data. But I think you'll see more diversity. If you look at Novartis, for example, Novartis is very invested now in the idea that human proof of concept through genetics is incredibly important. They're betting more, in my understanding, from talking to Mark Fishman and others is betting more on rare genetic diseases where they get approval in small numbers of people and then extend to a bigger market. So if you take, but that would be equally served by such a system. Again, we shouldn't know. Again, it's not that we're saying it'll be a GWAS system or a Mendelian system. It would be whatever question you had that company. Look, I think you get a much broader uptake if you make it clear it's inclusive in that way. David. Yes, sir. I'd just like to make a comment with respect to this just to put things into perspective of what percentage of people in the pharmaceutical industry know anything about biology. And it's less than 1%. So this is why you're not going to get the biological understanding inside the pharmaceutical industry. It has to be people outside the pharmaceutical industry. That's the first point, actually. That's the only point. Except, Jeff, I wanted to ask you a question. How important do you think the granularity of the phenotype data is going to be to the utility of the database? So from our perspective, we think it's very important. So right now, if you look at our external collaborations, we have access to about 100,000 individuals that have been followed over population studies, cross-sectional studies. About 20,000 of them have had metabolomics work done on them. And that's what we prize the most. It's that ability to complete the picture. For this particular variant, what have you, is there a metabolic fingerprint associated with that that we can pull back to a particular pathway? And so the idea of being able to call people back, that's very appealing to us. Because we have collaborations now with a number of groups that they do metabolomic type of profiling on a wide scale. And we take advantage of that currently. Please. I don't know how to ask this without sounding critical. So I'll apologize in the beginning from the start. I guess given Pfizer's history the last five years, maybe 10, and given the recent experience with CETP inhibitors, I'm really surprised that you're interested in triglycerides, insulin, fatty acids, and not stroke, heart disease, kidney disease, peripheral vascular disease, and heart failure. So who cares about glucose? You care about dying. Yeah, I understand. This is not meant to be inclusive at all. This is just a snapshot. But we're absolutely interested in those areas as well. I just want to ask a question. Perhaps it's not part of this discussion. But one of the unique things still in industry is your experience as you move the target down into large animal biology. Most of the academic community is still largely involved in studying mice and rodents, really. In just improving the future biology, so to keep the circle moving, do you envision a point in time where that biology is also made pre-competitive and made available? So that we know what the next generation of effects of human phenotypes are? Yeah, so what I can say is that we're entering into more pre-competitive partnerships with our pharmaceutical peers, where we share that type of information. And I would say we're doing that on a more limited scale with our academic partners. But whenever we enter into an academic collaboration, we basically give up all IP. The only IP that we care about is the chemical matter. So when we do find an academic, and that's new for us. I mean, that might be a surprise to many of you. We have no aspirations of owning the IP around targets and pathways. It's all about the chemical matter. And I think as more academic collaborators become aware of that fact, that door will open up. But right now, there's a lot of folks in academia that just don't even consider partnering with us because their past experience has been that it's such an IP nightmare that it's not even worth the effort of the 24 months that it often takes to get a contract in place. Just curious, when you say chemical matter, does that include biologicals? Yes. But in terms of the biologicals, we're not going to try to assume for a moment that we could protect a GLP or something like that. But it's the novel construct, the way that we would extend the half-life for the potency or something like that, but not the protein itself or the biologic itself. Yes, please. I was interested in to what extent already available phenotypes that are being generated in the medical clinic would be useful or usable to you as opposed to these very careful research-driven prospective types of targeted phenotyping. Yeah, I mean, so I can say with Duke and their CAFGEN study, there's a study where they're looking at individuals that keep coming back to the CAFG clinic because of a variety of different risk factors. And so that's a data set that's very interesting to us because it has the GWAS, it has exome sequencing, now it has metabolomics, transcriptomics, proteomics. So those are the kind of studies that we're constantly looking to have access to. We're now just starting to engage in additional conversations with hospitals and clinics. Could we get similar types of information? In fact, with Tufts, we just started a study this week where we're going to have access to samples from individuals coming into the gastric bypass in the banding clinic. And that'll be evolving over time and we'll continue to get samples both for genetic analysis as well as getting adipose tissue to do cell-based assays as well. I guess what I was asking is phenotype data that's being generated anyway through the medical care system as opposed to these highly expensive phenotyping activities. Yeah, so I don't have a lot of those conversations but I do know that folks in our business development and our external research groups, they struggle with that because of the consent that's in place. And so when the pharma company comes in and says you're doing something of interest or potentially of interest to us, we would like to have access to it. There's a lot of consent issues that do come up and in the question always is, is this big bad pharma trying to get some information that they're not going to share with the rest of the community? And we have to figure out how to work through that. But we have not done that very well at this point. Thanks very much. Okay, thank you.