 It's my enviable, I'm sure, pleasure to be able to try to recap what it is that we discussed yesterday. I actually have had a lot of help from the planning committee, so hopefully we've been able to cover a lot of the activities. I started yesterday by listing a series of questions we wanted to address, and there's actually some still outstanding questions that we haven't gotten to yet that we'll be addressing today. But the first question was, do we have adequate data sets and accessible databases to provide the data on genetic variants and the evidence supporting clinical actionability? Now, you'll see in a minute, I think we're going to need to have a debate about whether we ban the term, I see David's not here, but whether we ban the term actionability and return to clinical validity and clinical utility, and actually it might be very interesting to have folks comment on that as we proceed during the day, and I'll come back to that in a second. But the answer for whether we have the right resources, of course, depends on the audience. It's clear we have multiple audiences in just even within this group, ranging from the people that are actually producing the data and the associations between clinical variants, between genetic variants and clinical annotation, if you will. We have genetic laboratories that are fairly sophisticated in using a variety of resources that are available. We heard presentations from several of them yesterday, and you could see they were going out to a variety of resources, integrating them, and in some cases even writing your own software to do that. And then we have sort of, and there's others, but at the other extreme, we have the practicing physician, the primary care doc, who we imagine is going to be somebody who wants to use this, and everything in between. So it really depends on the audience, and even for a practicing clinical geneticist, I'm going to report something that Gail Jarvik said to me this morning. She said, I want to just be able to feed my DNA sequence into something and have it spit out the four things that I need to talk to my patient about. So that really sets the scale of the problem. We think Ensemble and ClinVar that we heard about yesterday are really good starting points, but they're probably a layer below where Gail wants us to be, or maybe two layers below where Gail wants us to be. Primary care docs clearly need at least that level of support. We think that there's an EHR integration layer, decision support layer, and when I refer to EHRs here, I'm referring to a generic EHR, because whatever form they take, they're going to have, I think, an important role going forward. We need much more clinical annotation. This is something, I think, one of the things that we're struggling with is, even though we've got massive amounts of data, we just don't have anywhere near enough we need to actually give us a full sense of the situation. And we especially need, I think, a place where we can capture variants of unknown significance. So at least they're recorded somewhere that this has been seen, and we don't know what it means, so that it can either be focused on by people that are making associations or it can have data added to it later. We need a mechanism to capture these one-off associations that we've heard about, where a laboratory is doing a clinical sequencing project, they have a focused clinical question they want to answer, but the value from recording that, I think, is very large in terms of the biology that underlies it and how the research community might be able to use that. And we think that the database needs to really carefully model classes of evidence, specificity, sensitivity, prevalence, PPV and NPV, penetrance, and there's probably a variety of other things, but these are really key elements of what the system needs to be able to capture. So, a tentative answer to question one. Any immediate reactions to that? We don't have time for a lot of discussion, but I'm just sort of interested in the immediate reactions. About right. Okay, terrific. Second question to consider is, what criteria need to be met in order to consider a genetic variant, or a pattern of genetic variants, clinically actionable slash clinically valid slash clinically useful? This I think there was less agreement or consensus about. I'm not sure we even really got to the details of it, but I think one point that was made is we really need to resolve the language. Language really matters here, so I think we need to come to consensus about, are we gonna stop using the word actionability, or does actionability mean something different from clinical validity and clinical utility? Perhaps actionability is something even beyond clinical utility, it's about actually doing the thing that you think is now clinically useful and clinically valid, but we can leave that for further discussion, but I think we do need to resolve that because as long as we're not clear about that language, we're gonna have some debate on what we're talking about. We, I think, saw good agreement that there was value to the use of bins, especially to the extent that it facilitated the commonly used phrase yesterday, low hanging fruit, the things which were easy, the things that, to use the schema we were using yesterday, things that fit into bin one are obvious, we should start with those. I think we also heard that a key driving principle in this needs to be the simple idea, do no harm, so if harm is to be done potentially by what, by use of any of these variants, that would be a good place to say, we're not gonna use them. We think we need to be willing to experiment with the bins of possible validity, so this would be bin two, probably B and C, again using the nomenclature from yesterday. We need to experiment, and I think there were two lessons that we heard. A, we need a robust methodology to actually think about how to experiment with it, and the other theme that we heard yesterday is we need to be able to be disciplined about discarding things that prove to not measure up, and that discarding is something that I think was a strong sense that we haven't been as good about. We need to avoid bins with no validity, and again, I think this is an area where we just need a ton more data, and we need to develop plans to how to think about addressing this whole clinical utility question. So I just wanted to pause for a second. I went on the internet and went to the PhD Foundation, not for any reason except that was the first one that came up in Google when I searched clinical utility and clinical validity, and I think everybody's familiar with these definitions, but I thought it would be useful for just a second to everybody to read the definition that they use. Validity is the accuracy with which a test identifies or predicts a patient's clinical status. For genetic testing, the relevance of a particular gene to a disease can be assessed by genome disease association studies, and the accuracy of the test is evaluated in terms of specificity, sensitivity, PPV, and NPV. Think it's a fairly standard, widely accepted definition. We need to think about, is this really the one we want to embrace? Similarly for utility, clinical utility was defined as an assessment of the risks and benefits resulting from using a particular test and the likelihood that the test will lead to an improved overall outcome. I was actually quite pleased to see that nowhere in the definitions did it require a randomized control trial. It didn't talk about methodology, but I think those definitions are helpful as we think about going forward, and we should have some discussion about whether we're still comfortable with that approach, and then this final question about, is actionability something even more limited than utility? And can we come up with a definition that we would all agree to? This is a question I didn't pose yesterday, but I think is really important since this is an NHGRI sponsored meeting, and I would argue the same, I would hope the same will apply to Welcome Trust, and I apologize for not including Welcome Trust in this. And we think the answer to that question is, we need to ensure that the discovery of disease gene and gene drug association continues through funding initiatives. This is squarely in the purview of both Welcome Trust and NHGRI, and it's not done. We have a lot more work to do in this area, so that needs to be a high priority for the funders. We think that NHGRI can serve as a convener in conjunction with other NHICs, in conjunction with Welcome Trust, professional standards organization, to foster discussions on clinical validity and clinical utility. To get to the point where there's some agreed upon standards for some of these variants. NHGRI, Welcome Trust, don't need to be the deciders. We're just suggesting they might actually be an appropriate convener to get everybody around the table. We also heard some challenges, again, in this boundary between research space and clinical space. And I don't know that if we went around and surveyed the room, I'm sure we'd have quite a bit of fuzziness about where each of us chooses to draw that line, but it seems that maybe having a discussion that would involve OHRP and other appropriate organizations, especially in light, at least in the United States, or things like the advanced notice of proposed rulemaking, might actually help us with some of these boundary issues. And then we need to create support, a resource for clinical annotation that extends beyond Ensembl and ClinVar. We think, I think Ensembl and ClinVar are a great starting point. But if anything, what needs to happen is there needs to be more clinical annotation. And much of that's a pretty manual process, although there may be ways to facilitate automating it. So that's another important role that NHGRI can think about, is how do we produce that layer that needs to sit above the ClinVar and Ensembl that Gail can feed, or any of you can feed the sequence of interest into to get results. So just want to, again, take a second. Does that seem about right to yesterday? Did I miss any key themes from yesterday? Amazing, that's not possible. All right, so we have a lot of work left to do today in our two thirds of a day or so. And so, again, I want to reiterate the questions that we're going to pose today. So, and then hopefully at the end, we'll summarize what we learned from those questions. So, what is necessary to integrate those data sets and evidence into electronic health records and into clinical use? What decision support and physician education will be needed in the clinics? How do we create a dynamic loop that recognizes the anticipated rapid increase in available evidence and upgrades clinical action ability, validity recommendations to actually doing something? And I think the hope is that at the end of today we'll be able to do for those questions what we did for the previous questions and add to that list of what NHGRI should do, what Wellcome Trust should do and what we would recommend others to do. One teensy comment. Since we have many NIH institutes here, perhaps we could say what NIH should do because we could be the convener, but really these are disease specific things that we would have to do in partnership. Yeah, and if you notice on that, it did say in conjunction with other NIH ICs. Yeah, I think there might be a little misunderstanding about the level that ClinVar sits at. It is the level of clinical validity, so it collects and aggregates the assertions about the clinical validity of variants from lots of sources. So it's kind of already one step, so that could be encouraged and developed, but the other is I think it is correct that review of that validity has to be done manually. And so recommendations could be built on that aggregate and that would be sort of a new layer on top. So it's just to keep sort of those things discreet. Yeah, great comment and it would be terrific if in fact ClinVar were the place and all we needed was more clinical annotation attached to that, that would be terrific. It also seems to me that it's necessary to have some regulatory clarity about what elements of this process are open to traditional regulatory process. For instance, if we create an automated tool to sort through some of the data from a regulated instrument, is that automated tool also a medical device? Great comment. Yeah, power. The risk of taking us too far down a rabbit hole, I think that starts to raise an interesting question as to whether these types of guidelines and recommendations and tools should just be the purview of public government bodies or should private commercial interests be allowed or for that matter encouraged to develop and market these tools on their own and how to what extent does that need to be regulated? How much transparency needs to be required? Where's the line between protecting the public and allowing free enterprise? I think it's a great comment. And a lot of us have thought a lot about that issue in terms of whether these kinds of resources should be allowed to be commercial or governmental and it's a great open question. I think one of the key issues that has driven much of genomics, at least to date, is open access and openness. And so I think that's gonna be a challenge in the commercial sector or how to deal with that. All right, any final comments? Not final comments. Intermediate comments at this point. All right, well then, I will turn the podium over to Tim Humbert from Sanger Institute, who's gonna moderate the next session. So everybody's coming up on.