 I'll skip that. So in thinking about kicking off the discussion, I thought about obstacles and opportunities. I mean, this isn't the first time scientific community has thought about large-scale sequencing and cohorts. And so it has a history. So I thought we could think about what are the obstacles and how can we overcome them just as a way of kicking this off? I think Rick had a real nice slide toward the end. Sequencing costs are high, but they are coming down. And I don't think it's a good idea that we make an argument that we should weigh. I always harken back to the old PC days, as when should you buy a PC? Going through the XT, the AT, the this, that, and the other thing. And it's going to constantly evolve. And we're going to use the best technology for the question at the right time. Peter talked a little bit about this. I'm going to say more throughout this discussion about the IT infrastructure. One of the things that concerns me is the current biomedical, at least, biomedical IT infrastructure really isn't up to the task. And I'm extremely excited about this project to actually enable it. And being the catalyst that enables improvement in the IT infrastructure in the biomedical community. I think one of the things that concerns me is if we create, and I'm going to use the word scientific commons here in a couple slides, if we create this large space where we have large-scale phenotype electronic medical record and genomic and other omic information that's linked in a way that it can be adequately queried to answer questions, that there's only a few places in the country in the world that can actually handle that. And I'd hate to see genomics and biomedical research go the way of physics, that there's only two or three places that have the collider appropriate for the experiment. So I'm wondering if during the next, whatever it is, 16 hours, we can think about the use of the cloud. As an example, I'm not pushing the cloud, but use of creative IT infrastructure. So we not only are making these data available, we're making the data available so it can be used to answer questions. I think, you know, and it's not just a server that's serving out data to other people. It's actually a place where people can query the data to answer questions. Another comment I hear a lot about is the perfect study samples not available. You know, I don't think there is gonna be any one perfect study sample that's ever gonna be available, frankly. And but I think we need to think about ways that we can bring together collections of existing and emerging samples that will achieve the goals that we come up with over the next few hours. And stop thinking about the perfect study that we're all looking for as we can bring together. I think we come very close by bringing together existing and as I'm emphasizing there in emerging studies. Another is that, you know, this shouldn't be seen as a U.S. effort, both this meeting and I actually, I think another exciting opportunity is to rekindle the international spirit that the genome project had. And we can, this study should reach around the globe. I know there are initiatives being launched. I don't think just in NHGRI, but across the NIH in Africa, for example, obviously there's a lot happening in Asia. So I think this large scale cohort studies, you know, sequencing in large cohort studies should be seen really as an international effort. I think it's a great opportunity. I guess sitting in Washington, we have to address that this really isn't the right time. You know, people talk about, is this the right time to launch such a study because of the economic uncertainty in the country? My prediction is first that this kind of study will be done and if we launch it in an efficient and organized way and we're result oriented, at the end of the day we're gonna save money by giving it the structure of people around this table and again being very driven and being very efficient and not just piecing it together over time, which I think in the end would cost a lot more. So by creating this resource is the word I use for the scientific community in this commons, I think at the end of the day we'll not only save money, we'll promote better science and better healthcare. The other thing is that institutional and disciplinary silos are gonna undermine this effort and I talked about it at the other meeting. I think we really need to work hard to remove those silos and make sure that across the scientific community and the epidemiologic community, the genomics community, healthcare, the basic sciences, the leadership works hard to bring all those people on board, that we do a lot of community engagement, in this case not the community of the participants but the scientific community and make sure they're fully engaged and see the benefits of such an expenditure and try to think of ways from them and within that these data can be used across the scientific community. Whoops. Freudian slip. And I mentioned the scientific commons and I just threw some large studies out there just as examples. I think we need to think of ways where multiple of these studies, large studies, and I put the charge is a lot of NHLBI cohort studies coming together, UK Biobank is represented here, a large initiative of the National Children's Study, I think contribute, we should not think of this only as a study of the elderly but across the lifespan and there are many large cohorts that can come together and we should talk over the next day of what the scientific commons looks like of bringing these data together as Peter said, that has both genomic information, phenotypic information and it's growing and increasing in value over time because of the accumulation of phenotypic information and our ability to recontact and remeasure people have specific mutations that we've identified. And then I've already touched on two different analysis models. I think the typical DBGAP kind of models is this commons is really a data server and it's pushing data out for other people to analyze and we can argue about how well that works and it has been argued how well that works but really I think that's one model. What concerns me is this data is gonna be so large and complicated that's quickly going to outstrip the vast majority of investigators and I think we need to come up with more creative ways of which people on the outside are reaching into the scientific commons and actually analyzing it centrally and whether it's in the cloud or not I'm not an expert to argue but I think we need to find ways of which investigators across the spectrum of science can analyze these data when we create it. It will be an outstanding resource but if it's sitting somewhere and they can't download it and they can't work with it in their home structure it's actually even though we can tout that it's publicly available it's useless because it would overwhelm them so we need to create opportunities so they can ping the data from outside. And then I also think there's ways of organizing this. I've got several ideas. One is think about this as a cross the lifespan. I think one of the things we tend to think when we talk about sequencing and large cohorts and well represented around this table we're typically thinking of the elderly people with complex diseases of later life, cancer, heart disease, diabetes and the like but I think it's really wise for us in terms of engagement of the community is there a complex diseases throughout life? The autism's been mentioned several times. There are numerous complex birth defects and I really think when we think about bringing together and the criteria for cohorts to join us I think it would be a mistake to think about just later life. You'll also notice by the way that there tends to be a dearth of samples in the middle years. That could be because there's a dearth of disease in those middle years but we really should as we bring this together think about whether that needs to be filled in. And I'm repeating myself now we have to make sure that it's representative across ethnic and socioeconomics strata. I won't say more about data sharing that's already been hit. I think the other issue is we need to do better of recognizing our successes. We're trained to be critical and many of us are well trained to be critical but on the other hand we have in this community a large number of successes and we need to get them out to the community and make sure we tout the successes of genomic research and population based research. It has offered medical care a lot and so I think we need to, as we build this it's not gonna be built overnight but as we build it we need to think about the successes and get it out there. Then finally what we're coming together is really, it's quite a large ideal that we're trying to reach so I would encourage us to set big goals. I'm an optimist and I like big goals but at the end of the day we have to break those big goals into a series of defined steps and timelines and deliverables and the slide on the left is I think we've met the enemy and they're us and it's up to us really to well articulate those goals, bring the community along and define what are the steps to reach those goals. So that is my two cents to kick this off and the floor is open and I encourage others to contribute. One, one, one, one, one, one, one, one, one, one. We can put some comments on Eric's talk. You wanna go and get plugged in and that would be great. Thank you. So on this issue of the adequacy of some aggregation of current patient resources. Yeah. It gets back to the question that we discussed a little bit the first round about the nature of the phenotypic data that you want and that's where I actually favor this rather first pass of broad phenotyping for some of the reasons that Francis articulated. That's where I see some weakness in the currently available resources. I don't know so much about some of the non-US ones but here the strong incentives operating primarily in the biomedical research community are to get a collection of patients that were chosen on rather narrow phenotypic grounds and to study the relevant aspects of that phenotype in great depth and I'm just not so sure that if we put lots of those things together we're gonna get what we need. I would agree with you but also there are sample collections of very deeply phenotyped individuals that have literally thousands of things measured on them across disease entities. It is true they tend to have a disease focus but they're not simply cases that were sampled from a clinical setting for example. They're very deeply phenotyped and they have been very deeply phenotyped over time and I think many of those would at least partially meet your criteria and then going back to the previous discussion I'm a big fan of once we then invest in the sequencing and we identify homozygotes for example with PCSK9 deficiency then you can bring those people back in into what we used to call CRCs, the CTSA setting and even more deeply phenotyped them in a clinical care setting. So going that route we'd presumably have to prepare for a lot of reconsenting and so forth. I just don't think that there are very many groups out there that are consented in a way that will fuel your scientific comments and so forth. I would guess there's probably 60 to 100,000 people that are well phenotyped that are consented for recontact. That's a ballpark. I'd like to hear more about them. I might just comment on a workshop think tank that we had held a couple of years ago asking the same idea, sort of new models for prospective studies and recognizing that many of the diseases, the incentives are or the cohorts, there is an incentive to focus on a specific disease but wouldn't it be cool if in each of those cohorts we could collect some core of information that is common or is maybe not common but is critical to all of the NIH institutes. So that in the cardiovascular cohorts you do collect cancer and mental health and not a lot of it but at least enough so that you can do some basic phenotyping and some basic evaluation of this course, pick out some interesting cases. And so that was a model that was proposed. I don't know that it's gained a great deal of traction but it may be something that we'd want to consider over the course of tomorrow. I hate to say this, can I bring up a point on that topic, I hate to say this with the gentleman to my right is why I think one of the issues we have is we tend, we have these institutes under the NIH and they have disease with the exception of a few, they tend to have a disease focus and the money goes through those and so we tend to have diabetes cohorts, heart disease cohorts, cancer cohorts, et cetera and that's one of the silos I think we need to, not to bring down but to restructure so we can phenotype across the NIH in a constructive way. And I think that's happening in the presence in the room of representatives from many of those ICs is just I think a very concrete representation of that recognition that the more we are doing things in a trans-NIH way the better we all are downstream. I think it's unlikely we're going to change the structure of the 27 institutes and centers given how much pain and suffering it was to change just one of them over the course of the last two years. So what you see is what you get but we can clearly work along the lines that you're talking about and I think there's a lot of motivation to do that. There was a click. So to follow up on your comment I think we all agree on the importance of longitudinal follow up of folks. I mean, how do you actually see that happening? Is it going to be through the research unit, through the healthcare provider, through patient-based or all of the above? Practically speaking, how do you describe models where that's worked? You know, the kind of cohort, prospective cohort model is one that's well in place within the NHLBI. There are a number of cohorts, Framingham's one I'm familiar with but you know where it's set up and you've established biobanks and have a lot of things in place to be able to go back to either samples or the participants and have consent in place. So that's sort of what you might call the prospective cohort model and maybe Rory can comment on the kind of UK biobank kind of model. Can I say that essentially the same thing is we'll talk a bit more about it tomorrow but what we're calling cancer cohorts, you understand, don't start with people with cancer, they just start with people. Two million of them enrolled in 47 cohorts and last year we pulled in a third of the cohorts in the club, pretty extensively go after a couple of other endpoints too. Not everybody does everything but what Dr. Collins was describing is I think what everybody who's got a cohort is trying to do, that's one. And then I would say and these are older cohorts, it's a mix of young and old but I'd say maybe a sixth of them also do have routine sequential re-contact. Some of the contact is as far away as 10 years but for the cohorts that can afford it then they go back much more frequently. So these things are pretty much what cohorts like to do, want to do, money permitting. I think it's done more uniformly in the biobank and maybe you might want to talk about that. I think in general one wants to embed these cohorts ideally in a setting where you can get follow-up of the widest possible range of health outcomes easily. You want to do it on a large scale and that was the basis for UK Biobank so we can link electronically to death, cancer, all hospitalizations and from the end of this year all primary care. So all of those data are coming in and being linked into the database and we have consent to then go back to participants to get more information about the health records but I think the general approach would be to embed the studies in situations where you can do that readily if you want to do it on a really large scale. I'll just echo what Rory just said. I think that to do this right it makes sense to link to health care outcomes and so my task tomorrow will be to talk about the electronic record and those sorts of issues but the electronic record is totally agnostic with respect to diagnosis and interesting things happen when you start to excavate in that set and it turns out that it's not really simple to excavate from that set but I think that that's another really, really promising approach. And I, Eric, I think you ought to add electronic record cohorts up there. I'm looking at my numbers and the Merge 2 has 313,000 samples. Or 313,000 participants, yeah, samples. Comments? So we have had, people will help me if I miss someone too. Individuals have come in during the discussion. Nancy Cox and also Gail Jarvik, so welcome. Did I miss anybody? Snuck in. Probably saw people sneak out.