 While he's getting these up, I'm going to take a much higher level view of the day and of the challenges. It seems to me, whether it's a merge, whether it's PMI, whether it's the MVP project, really what we've been focusing on is two transitions. The first transition is how we use EMR data to make discoveries. Taking healthcare data and making novel discoveries. The second transition is how we push those novel discoveries back into clinical care. Very often in our conversations, and this is not a criticism, it's just human nature of who we are as scientists, we tend to focus on one or the other. I think the challenge for all of us, both individually and primarily by the way, this is an institutional challenge, is how we create what I call in this slide a virtuous cycle where healthcare data is used to fuel discovery and with appropriate vetting that those discoveries then are brought into healthcare. And if we can create this cycle both institutionally as individuals, I will claim that we're going to have much more success in our enterprise. I also took a higher level view of my assignment to give a response to this idea of evidence. And you'll see in some slides coming up, particularly picking on and leveraging ClinGen, there's kind of three kinds of evidence that we're all involved in. Most of us are very familiar with sort of the discovery evidence. You know, we're P-values and replication, FIWAS, etc. We're very familiar with that circle. I will claim though that there's an emerging challenge in that circle with N of 1s, particularly in the clinical space. We don't know yet as scientists how to handle when we have that very unusual patient with a very unusual variant and as Tom Kasky once told me, you know, Eric, I've never seen this before, you know, is how we interpret that, you know, and these N of 1 trials I think is extremely challenging for us and it's a fun challenge by the way. But the other is you'll notice that experimental variation is often brought in. If we wanted this a perfect experiment, that would be that N of 1, but really once again there's P-values, replication and whether we admit it or not, we all are biased by prior expectations. On the translational slash clinical side, it's more interesting, I will claim. First, clinical impression is absolutely entrenched in the field and at least, you know, for the coming years we're not going to change that but we shouldn't change that by the way. But clinical impression is very entrenched. There's also professional standards, there's experts, there's societies, you know, there was 48 acronyms in the previous session that are helped developing those standards. And the other is that, you know, the field doesn't like contradictory data and Gail was using, you know, the word penetrance several times. You know, it's a fact of life that there is variable penetrance and variable effect sizes. So it would be interesting to see, for example, all those people that have those cardiomyopathy variants, you know, they don't have a diagnosed cardiomyopathy, do a good echo and see how large their ventricles, their wall thickness and their injection fractions are because it's a continuum. You know, just like the penetrance is a continuum, the clinical presentation can be a continuum. So, you know, and this has been, you know, you all know this better than I do so I'll go over the next few slides very, very quickly. This has really been ingrained then in a series of papers and, you know, in ClinGen. And I only circled, you know, first of all, even in the cheap seats in the back, you can kind of see this definitive strong, moderate limit, et cetera. And also then there's the ingrained idea that we can use. There we go. You know, we can see this, there's basically this idea of replication, bringing in experimental data, and then once again, that not liking contradictory evidence. These are all brought together then to help various people, you know, adjudicate and put variants into categories. And Heidi Rehm very generously gave me these couple of slides, you know, and this has really been ingrained into a wonderful resource in ClinGen where, you know, groups of individuals, sort of domain experts are coming together to create, you know, evidence that can be used both for clinical validity and utility. I'll just pick on those two in actionability. And once again, it's not just, you know, it's fun for me to go through this. It's not just using information from one domain. And we shouldn't, for those of you who are not in this space, you shouldn't underestimate the sheer amount of work it takes to bring this information into a space, have a group of experts around the phone lines, and coming up then with recommendations for the community. But I do want to make a few comments about actionability because that's been a major part of emerged activities. And once again, I'm not presenting the tables from the paper, but I thank some of my co-presenters calling this Hunter paper to my attention. And once again, this, I would say in terms of areas of difficulty, this even went one step further and brought in this idea of severity, the effectiveness of the action, the nature of the intervention, and once again trying to come up with a score which is no easy task. But I do want to make a couple of editorial comments that usually, you know, actions usually considered a modified treatment or preventive measure, and I want to emphasize sometimes the preventive measure idea of this, that you can often use this information to prevent the onset of disease, not just to change the therapeutic track. But I also want to make the point that even reporting the information is an action of itself. And there will be downstream responses to that action that we do need to take into account, whether it's the family, family planning, et cetera. And the other point is that there's a tendency in our discussion to think that the action, that misapplied action will have very detrimental consequences. And we weigh that. We weigh that whether to apply the action, whether to put it into the bin of actionability or not. But in fact, many actions would not have great detrimental consequences. And that, again, should be taken into account. Just to be provocative for this group, I picked on FH and LDL receptor mutations. It's interesting to think whether projects such as Emerge, and we've heard about today, whether physicians will begin to treat the genotype and not the phenotype. Again, I picked on this one because it's fairly easy to understand. Physicians today treat your LDL cholesterol levels, and they treat them by giving you a medication that typically works to lower those levels. It's going to be interesting to see as the field matures and this area becomes part of routine care, which is supposedly our goal, whether physicians will actually ever treat the genotype without the onset of LDL cholesterol. That could be done for preventive measures, number one. And number two, there's no data whether people with these mutations but do not have high LDL cholesterol, are they at higher risk for having an MI? Because of perhaps the integrated effects of those mutations both biologically and over time. So these are things, again, to think about. For those of you who can't see this, what we've seen so far and up to this point is great, but I think this is line is why we're worried about the future when the present is more than most of us can handle. But what we've seen so far, it doesn't scale. This is Ethel Merman and Lucille Ball and their famous chocolate factory scene from the Lucy show. What we're doing right now, it's great science and in fact it's great clinical care in specialized arenas but it will not scale. So we need to really challenge ourselves of how indeed we can get these to scale. And the reason we need to scale is basically there's just more and more and more sequencing done. And I just picked on these particular studies because they're doing a lot of sequencing. And so both within the clinical arena, both within the private sector and in these very large research areas, even today in 2017, there's an enormous amount of sequencing being done in the US, in Europe, and in China in particular. And so we need to think about taking these emerged discussions and making sure we can do them at scale. And this is a slide that Richard Gibbs gave me and it's a collaboration with Richard and Will Salerno and Eric Venner that's here, is the idea of having semi-automated clinical reporting of building a machine or a pipeline, really not a machine, building a pipeline that will help with clinical reporting to help scale this process. This slide's already been shown but I'll show you, when you apply such a pipeline to the emerged data you get about 3.5% actionable variance. But I do want to come back that even a pipeline like this still has a lot of manual intervention. And that's where I'm putting my attention there is how we can, I don't know if we want to get around it but how we treat that manual intervention and how we scale it given the time that it takes. I think it's something that several speakers today have talked about. And there's a lot of ways to think, I'm going to sell crowdsourcing. And in some ways, ClinGen itself, it's a cloud source, in some ways it's a group source and maybe we can take something like ClinGen and put it on steroids and think about crowdsourcing. Think how can we leverage the very large numbers of individuals that are in our community and bring them into this fold to help adjudicate variance, for example, that basically help with that red circle in those purple boxes. Because again at the moment, right now if we're going to take genomics into routine health care we're at the point where we really need to look at this scaling factor and begin to address it. So Sharon, those are my comments. Great.