 So, we had a spirited discussion. All of the usual suspects were there in the pharmacogenomics arena, and I'm sure they'll chime in if there's anything that I've missed or misrepresented. Our understanding of our charge to the working group was that we were to discuss and identify one or more collaborative demonstration projects that would advance implementation that should be a capital I of pharmacogenomics into clinical practice. And I have to say, in our discussions, as we talked about applying sequencing approaches, the discussion each time moved from implementation to discovery, and then we had to move back to implementation. So I think that was a recurring theme for much of the discussions. But there were some items of consensus, and they're summarized on this slide. We would suggest that perhaps more than any other field in genomic or personalized medicine, pharmacogenomics has actually advanced to the point where we're ready to implement pharmacogenomic testing into clinical practice. So in terms of readiness for implementation, pharmacogenomics may be more advanced than some of the other areas. And that's because we know about specific actionable variants, and in many cases the mechanisms by which they act. Secondly, pharmacogenomics is less shrouded, we believe, with some of the ethical, legal, and social issues that come along with some of the other areas of broader aspects of implementation in personalized medicine. I think we can discuss and argue about that a fair bit, but I think that that's probably a true statement. Thirdly, I think there was agreement that broad preemptive pharmacogenomic diagnostics is the direction we need to go, preferable over single variants, sort of reactive pharmacogenomic testing. That being said, there are some clinical circumstances where potentially single variants, reactive pharmacogenomic genotyping might be more effective. Perhaps the CYP2C19 Clopridigrel story being won. Next, it was generally agreed that there's a need for a coordinated effort to develop best practices for implementation and to develop a framework to advance new discoveries. And on a number of levels, including technologies for variant detection, defining which variants are actionable, sharing genotype-phenotype definitions to sort of standardize them in a better way, suggested actions, educational materials, local best practices, et cetera. And finally, coming back, maybe again, moving more to the discovery phase consensus that we really did need a centralized repository to annotate rare and new genotypes of unknown significance and, in addition, phenotypes connected to them. And of that bullet, you can see that there are already a number of ongoing efforts largely as part of the Pharmacogenomics Research Network to address each of those four sub-bullets. And we thought it would be a good idea to leverage these ongoing efforts rather than recreate some of them. And in particular, CPIC is addressing the defining of actionable variants and also standardizing genotype-phenotype definitions, suggested actions, sharing of educational materials, et cetera. The Translational Pharmacogenomics Project that I mentioned as well as opportunities for collaboration addressed some of these issues. And also the VIP Pharmacogenomics Sequencing Platform being developed and potentially a demonstration project that will involve the Emerge Consortium. So those were all items of consensus. One interesting paradox that kept coming up over and over again is that, at least for implementation, if Pharmacogenomics is a little bit more advanced in terms of readiness for implementation than some of the other fields, paradoxically speaking, genotyping is probably a better direction to go than sequencing. In fact, in some ways could be regarded as a far more elegant approach to implementation of diagnostic testing. And I guess the analogy I'll use is that if I'm in the clinics seeing a patient with diabetes and I want to know what their glucose levels are, I don't measure a metabolomics profile, I measure a glucose level. And so in the field of pharmacogenomics where we know genotypes, at least for implementation, genotyping may be a more elegant assay than sequencing. But that being said, if we all believe that next-gen whole genome sequencing will be the genetic test of the future, maybe the not too indeterminate future, we do need to know about how the pharmacogenomic tests or genotypes who are interested in what the quality of these are in a next-gen sequencing platform. And so reworded in a slightly different way in terms of a pilot project, a demonstration project. It would be of interest to perform a study in which we could look at next-gen sequencing and determine for implementation, do sequencing platforms add value over direct genotyping platforms for pharmacogenomics implementation? And the reason I think this is an important question is that some pharmacogenes are sequenced poorly on these off-the-shelf whole genome sequencing platforms. And we may need to develop next-gen whole genome sequencing platforms that cover these pharmacogenes better so that if there is a single genetic test that will be done, we have the pharmacogenes covered properly. And as I said, we know what the relevant actionable variants are, so why sequenced? So the proposed collaborative project could be to compare head-to-head, so to speak, whole genome sequencing, directed VIP pharmacogenomics sequencing, and low-tech chip-based or other genotyping platforms in an implementation project. And we could measure the technical aspects of variant calling and also potentially relate drug response and adverse events to these genotypes. A somewhat related project would be to ask what is the role of rare variants in pharmacogenomics implementation? And again, this is where we kind of blur those borders between implementation and research, but certainly a proposed collaborative project could begin to apply whole genome sequencing technologies for rare serious adverse events for both discovery but maybe also family-based pharmacogenomics implementation. And I think you've heard this echoed over and over again, pharmacogenomics is no different, but a repository for well-annotated rare variants in pharmacogenes would be very valuable to the field going forward, as well as not only annotating the variant, but also evidence for any functional or clinical consequences. So I think I've summarized most of the salient points, but I'd encourage those who were part of the discussion to chime in. So Alan has done his usual, very clear job of summarizing the discussion and his usual diplomatic job of hinting that we had a vigorous discussion. So I think he won't be surprised that I think he accurately described the outcome of our discussion. And I would respectfully offer a minority report, which is that genotyping is great, and we're all doing that, but where the world is going is sequencing. And I would prefer that pharmacogenomics be in the vanguard of that effort. Now he was talking about whole genome sequencing, and I see Debbie sitting across from me. And Debbie is not bashful, and I know she'll speak up in about two minutes. She's leading an effort within the PGRN, working in collaboration with both the Washu and Baylor Genome Centers to develop a capture reagent. That was this VIP thing, you know, I would have to find a better name than that, but a pharmacogen capture reagent for sequencing. Which in some ways may be a useful intermediate step, and I know that Terry's holding the microphone, but Terry will be able to speak for herself in just a minute, that she's thinking about perhaps applying through the eMERGE network. And frankly, I think that's a good idea, and I will immediately declare a conflict of interest. I'm not a part of the Mayo eMERGE project, but I think that that would be a good idea, but beyond eMERGE. So I'd like to think about something like this for pharmacogenomics beyond just eMERGE, and I hope I'm not saying anything politically incorrect. So everything Allen has said, I would agree with, with the exception of wanting us to move very rapidly toward sequencing technologies as applied to what I also believe is the leading edge of the application at the bedside, and it's already happening, it's been happening in many of our places for years, of genomic science for clinical decision making in pharmacogenomics, but Allen was entirely accurate on what he just reported. So I'll just speak up. I mean, the group has identified a set of genes and is developing reagents to be able to sequence these on a large scale in many, many individuals at a time. So it's a little bit different. What we're finding in looking at large scale or looking over large scale exome sequencing is there are many rare variants in these genes. Whether they actually have phenotypic effects, we don't know. The other thing is that many of these genes that we look at are worked together and they're not independently metabolizing or certain drugs. So knowing the full knowledge base of variation in these genes together, we may actually find some new insights. So I mean, that's what everyone hopes for, and to apply these to true clinical samples is really an important thing to think about. So I do think there are some great projects that can be done to prove principles that might only have one single SNP or variant. But I do agree with Dick that there is an opportunity with technology advances to do something broader. The concern there has been all of the incidental findings, if you will, and how to deal with them. And on that note, Greg Ferro and others at NHGRA have been engaging a number of the other parts of the community that need to be involved in. Wednesday Greg had a session, and Greg's here, you can comment further, but had a session with the major pharmacy organizations getting them engaged. And they're a group of people that is used to having to act on a pleiotropic way with data. So their drug interactions are basically a great example of how one event can affect a whole bunch of different pathways. And that's what often we're talking about in the case of pharmacogenetics. So I think some of the fears about doing broad genotyping may also help us get towards solutions because it will bring in people who frankly didn't even know there was a genome and are just waking up to that fact. This is a targeted panel. It has 50 genes on it. Well, it might be up to 100 now. But 83, excuse me, it's right. But they will be very well covered, and we're trying to cover some of the more difficult genes. I mean, this is an important thing. It sometimes is easier to cover these by whole genome, but then the incidental findings become an issue. So developing targeted pathways can help, particularly if they're really inexpensive to apply, which we hope long term this will become so it can be applied in any clinical study of a new drug. And by incidental finding, I don't mean accidentally looking at a gene. When we did CYP2D6 for Tomoxifen, we had to be ready to account for almost 25% of the FDA-approved drugs in terms of what I call collateral data, the collateral damage type data, where we looked for one thing and had to be responsible for all these other things. So even when you do it on purpose with just one gene, there can be a lot of information. I wanted to ask, are any of the places that are doing sequencing, actually reporting out pharmacogenetic information when they're talking to their patients and giving them actionable results? So can, is anybody? We are. The first comment I would make, Alan, is that when you order a blood glucose the test that actually gets run on autoanalyzers, 1,000 analytes, and you get reported back the blood glucose because that's what you order. The other comment I would make is I think that if you look at the economics of this, looking, we have to do two or three things simultaneously. One is we have to do what Debbie is doing for the PGRN, move the content for these pharmacogenetic genes onto a next-gen platform so that clear labs who have next-gen tests can provide that content through that methodology. But I think it's a denial of reality to think that sequencing isn't going to be the mode of analysis here. And for several reasons, I think it's important to be looking at this. One is the genome is not just SNPs. It's indels, it's copy number variants. It's the ability to look at microRNA and RNA-seq, and the genome is complicated. It's much more complicated in terms of clinically relevant variants than just SNPs. So that's the first point. And then the second point is you look at the cost curve. You look at the money here. We can't, you know, somebody who's running a clear lab in the context of clinical care, when you're providing content on a small number of genes that are well-curated, and you're providing that back to clinicians, the cost of that is greater than the cost of sequencing the whole genome. The information, it's true that the interpret, you know, it's the $1,000 genome, and as Bruce Korb said, the million-dollar interpretation. There's no question about that. But if we're all going to be on this platform, I think it's a mistake not to move in this direction now. Both from a research perspective and from the translation to clinical. So that's my feeling about it. And I know that, you know, a lot of us in the PGRN, there are few of us that sort of feel this way, and a lot of people that feel on the other side. But I got to tell you, from my perspective, I think that sequencing is the way to go here. So I think this discussion is very similar to the discussion we have in the smaller group, and that's good. It's a good reality check. And I think really there's more concurrence here than there is debate. And I think where there is the debate is how we define research versus implementation. When I think about implementation, I think what can we do tomorrow in the clinics? And for pharmacogenomics, I think genotyping is the answer. For discovery, I believe that sequencing is the answer. But for implementation in five years, sequencing is the answer. Right. And that was the basis of the pilot projects that we proposed. Well, I mean, I was there, and we had an interesting discussion from my perspective. And I think it's really important to say that for things like what Debbie's interested in, which is discovering rare variants, there's no question you've got a sequence. From the standpoint for which those of us that are trying to implement this and write algorithms to detect a rare variant that's never been described before or that I don't know what to do with doesn't do me any good as someone that's trying to implement. So I think that although there may be value to knowing that the patient has a rare variant in the context of saying, well, let's figure out what in the future we're going to say about people with this rare variant. That doesn't do me any good right now for making a decision on that. I don't think they're mutually exclusive. Not at all. And that's the whole concept. I mean, we know common variants will affect people commonly, and there are clearly examples of this. It could be genotype today to look at the effect of this. But you have to understand, if the objective is to actually implement, where it's to gather information that you can't use is actually a barrier to getting the study done. I mean, under current law, if you find any information in your study that is available to the patient's physician, that information is potentially disclosable to any insurance company other than a health insurance company that that patient may wish to do business with. So if you truly did inform consent, full informed consent, I don't think anybody under the age of 50 that might want to buy life, disability, or long-term care insurance, that matter under the age of 60, would ever want to be in such a trial, because that's the law. And it's just not going to change any time in the near future. So unless we have a way of doing whole genome sequencing and some way blocking access to the information for anything that is not part of the study, we're going to have a real implementation barrier. You know, I think that this is a different issue that's coming up here than the point of rare variation, which collectively can represent like 5 to 10 percent overall of the types of variations we see in populations. So collectively they're, you know, collectively they're common. Yeah, I would just like to follow up on what Scott said and maybe take it one half a step further in that I think what we've been talking about here are a lot of sort of tactical approaches to developing evidence. But I think the larger strategic picture is if there is consensus that whole genome interrogation should be developed to be deployed in the clinic, then I think all of these different facets of it, like pharmacogenetics and family history studies, should be designed so that each one of them is a brick that will be put into a structure that when you put all of the bricks together, then whole genome interrogation makes clinical sense, because it would be a complete failure if we only do this at the tactical level, and each study ends up saying, well, if you genotype the 83 genes that you already know are involved, guess what, genotyping is a cheaper way to do the same thing. You've completely blown it. So I think we have to do these kinds of studies in the context. Again, that's what we agree the goal is in the bigger picture of where we want to be at the end of the day. I think towards that goal, you know, we need to know that whatever whole genome sequencing platform is used covers those pharmacogenes. So that's the pilot project we proposed. So I want to make three comments very quickly. One of which is that somebody who volunteered for a research study had an echo and found out I have a heart lesion and I can't get life insurance, so this is not limited to genetic studies, and I was never consented about that issue because nobody thought of it. I think one of the things we need to be very clear of when we're acting like whole genome sequencing isn't in the clinic, it is. It's a reality. I have a whole bunch of patients that I have pharmacogenomic data on that we've got to decide is the pharmacogenomic data more accurate than the DMAT chip, which, you know, the problem with SNP chips is if you've got a variant next to it, it actually affects the quality of the cause. So I would argue that yes, there are problems with sequencing, there are problems with chip-based, there are problems with SNP-based. Anyway, you're going to genotype, there are problems with it. The reality is right now we can get reimbursed for doing whole genome sequencing, we can't for doing a DMAT chip in our institution, so we can do one but not the other, and the one we can do is the sequencing. And so we've got to get over the idea of which is about a platform and deal with the reality that probably increasingly most of our patients are actually going to have their genome sequenced. And so we've got to design our studies based on the fact that we're actually going to have that data, not, you know, pharmaceutical companies are now using whole genome sequencing in pilot studies. So this data is there. The question is are we going to use it, not which is the best platform for getting the data? I guess I was going to just echo what some points that have been made earlier, that is even if the project is clinical implementation and the vast majority of pharmacogenetic variants I'd be willing to clinically implement are ones that have been previously discovered and therefore are interrogated on arrays, I do think there will be some rare variants in genes that we know are super important, like CYP2C19, like TPMT, like CYP2D6, that if they were truly early stop codons or something, I would implement them. I know there's maybe disagreement among clinicians here about whether they would do that or not, but I think that's where we get to what Kate brought up yesterday. There's precedence for evaluating variants as they come up, like in BRCA1, BRCA2, and having levels of evidence where if it's really an early stop codon in a gene that's basically a monogenic gene for a drug that's important, even if it's the first time it's been observed, I would be willing to clinically implement it. Then you're going to have a whole other series of rare variants that you've discovered that you're not sure whether you can implement them in real time and they will go into the research side of the project. I think it is worthwhile considering sequencing in a clinical implementation project, even though I'm a DMET array girl. Maybe we should, just in the sake of time, since the next topic is sequencing, maybe we should move on to the sequencing workgroup.