 Yeah, so this is quite a bit of a different workshop from the last two. We don't have, unfortunately, we don't have a final workshop report to distribute to everyone today, still in draft form. Some of you who actually were at the workshop have a copy of the draft at this point to help with the discussion. So there's a problem which I think is widely recognized, which is that without enough functional information about genes, it becomes very difficult to interpret almost any result that you can get, any single result, any association result, any sequencing result, transcriptional, transcriptional data, anything implicating a gene for any kind of experimental study. 80% of genes have no experimental process annotation, and this is really critical. The lack of a framework for interpreting any individual implication of a gene is an impediment from moving to biological understanding and to translation. There's another way of looking at this. This is just a schematic, a cartoon of the gap between variation and phenotype. And it's actually taken from an old figure. You've seen other newer figures illustrating the various categories of research we have and programs we have and how much we invest in them. This is taken from kind of an outmoded view of that, but it does make the point that we fund a lot of functional genomics, and it's stuck somewhere here between variation, this gap between variation and phenotype. But the problem is that while we all talk to each other and while we're at each other, the information from this, these studies and from other studies that we don't fund, others fund, is not integrated and we don't have a good handle on what's really useful, useful way to integrate it. We don't have a good handle on what data are more useful than other data for biological and translational interpretation. So that's what this workshop is about, just to discuss some of these ideas. So what kind of a framework could they be organized into? Should capture and integrate functional data about genes and all the elements that regulate them? Should certainly include different kinds of functional information. We fund a lot of molecular information, molecular functional information about genes and related functional elements in the genome, for example, in code. We do a bit less of getting process pathway and partner data, but there is some, for example, common fund initiative called LINX. And ideally we want to include functional information about dynamics, cell tissue context, genetic background, et cetera. Again, it should provide a way of understanding what data are more useful and this notion of utility came up a number of times in the workshop for specific interpretations, so for translation or for understanding biological mechanisms. What data to include and what level of detail should be evaluated with regard to utility towards these functions? And it was suggested that to begin, again this is just a proposal, to begin at least one functional attribution supported by high quality evidence for all genes and elements. Another aspect of a framework that would be successful would to help inform priorities for adding new data types. So if you really had a handle on this, you could look at each successive data type that was added and think about did it really add enough value or you could look at pilot data and so this is really going to add to our ability to interpret these data. So there might be some way to get out of this way to inform priorities for adding new data types. And as you know, we come to you with ideas about adding new data types all the time. There are various ways you could imagine this could be done and this is one idea that came out of the workshop. You could have four elements and I'm not making any statements about whether these four elements end up as separate funded initiatives, that's for the future, but you would need four elements. One integrated, gets sort of gathered up, aggregated and begin to integrate existing functional data. The other function would be to evaluate the data and identify missing information. Another component would be to generate experimental data and finally, a final component, you have to know what quality you have of these data and gather data. You have to know what quality the data are for interpreting biological and maybe translational results. So you have to have some kind of modeling activity essentially as validation. It's difficult to summarize a workshop that had I think this many thoughts flying around because this is a huge area without leaving out other views and I have definitely left out other views and I just want to summarize two that I heard pretty clearly. There were certainly others that I think play into the scheme that was just outlined, but a couple that don't really fit. The first is if you're interested in understanding biology, underlying complex disease, you need to know about physiology rather than gene function pathways. There are a lot of different ideas wrapped up in that objection and I think all of us would sort of disentangle them in a slightly different way, but I think one of the most straightforward ways is to say before you can get to physiology you have to know a lot about the function of individual genes or it's much easier to get to physiology if you have that prior information. A second kind of objection that I heard was that a framework like this is too vague and we really should just pick something, something specific project and do it. For example, to understand how variation relates to phenotype we should begin with actual human variation and just look at the resulting molecular phenotypes instead of trying to do this for all genes or on a very broad-based way. So again, just to reiterate what would success look like, it would be a well-validated community resource of integrated functional data across multiple data types assessed for quality and utility. It would allow rapid interpretation for any result, implicating a gene or a gene regulatory element, would allow more rapid insight into biological processes and some cases would allow more rapid identification of points of intervention for disease and within an HGRI would provide a strategic context for the functional genomics programs that we undertake. And with that I'd like to thank all the members of the planning working group including Lisa, Shaila, Elise, Tina, Gatlin, Peter Goode, Mike Payzen, Ajay, Chris Wetterstand and Anastasia Weiss. Before going to general discussions I'd like to ask some of the folks that were there if they have any comments, especially Carlos. So I thought it was a really great meeting in that it opened up my eyes to how many different ways one could interpret genotype to phenotype and it really ran all the way from people who thought it meant we need to come up with a function for every gene in the human genome to we need to be able to create a resource to return results to clinical investigators. And everything in between, I mean it really was that broad. And so one of the comments that several of us made there was that it would be great to see this better articulate a vision that integrates with the roadmap and in particular how it relates to the transition from studying the biology of genomes to studying the biology of disease. So you know I think there were a lot of great ideas but going forward and thinking about sort of flat budgets, how to integrate that and ideally how it might integrate with existing projects like 1000 Genomes and GTEC and ENCODE and the kind of unifying role something like this might potentially play. Yes, so I had a similar reaction to Carlos. This is a very complex problem and it was a very heterogeneous group of people all with very different views as Carlos points out and we had a hard time I think coming to consensus on just about every issue that was discussed. I do think that in retrospect that makes me think that trying to tie this more specifically to a more targeted problem or use case or I think one of your slides mentioned that at the end might help to frame the initiative in a better way that indeed I walked out of their feeling that things were just too diffuse. And I think it was also just that the initial work document that they circulated was just you know it was a Rowshark text you would read into it what you wanted to read into it. Exactly. And that's why I think somehow to come up with a more targeted. But I think it's also just really critically important right I mean if we want to go down that road map and be returning results to people on iPads like what results are we going to return back and this is really kind of critical to making that happen. I mean I would agree that it was critically important and we all had a somewhat different view of what was important. Yeah Ross. Yeah so the suggestion that we try a pilot a focused pilot project that you know that actually that came up and there were some things proposed you know kind of the straw dogs and you know it was Katie bar the door I mean it was it's very hard to settle in on that pilot. But having said that you know I'm just joining in with everyone else saying yes this is we all recognize it's critically important and we need to find a way forward. I think that the pilot project is a good idea to clarify what could actually be done within this but it's going to be difficult but I still have to do it. Mike. I was just going to agree with what's been said and suggest that what this means is not that we wait a while to have another meeting. I think this is a place where a series of meetings perhaps with different specific foci but along the same general thread is really important because this is where we need to go. This is where NHRI is uniquely positioned to lead and the fact that it's hard to get there doesn't mean people aren't doing a good job. It means it's just hard and nonetheless extremely important. So getting people together multiple times in relative short order I think is probably important. Rick. I agree with I didn't get to attend I wish I could have but I couldn't. I agree with what Mike is just saying I think this is a really important function and I get the message that there were a lot of people from different backgrounds and it was hard to focus it but it doesn't do away with the problem. It is such an important problem that we're generating all these different data sets and encode being just one of many of those as well as the rest of the community and right now if you want to use it essentially you're talking about how do you interpret the genome right and it's not just variation it's where proteins bind it's where you know epigenetic changes happen etc. And more than that but which of those kinds of data adds most value for our ability to translate. Perfect and I think that the hard part maybe this is what was shown at the meeting too is that you're really having to integrate a bunch of different ways of looking and I like the physiology argument in the sense that you want to combine it with that but we are a genome institute that would figure out how to take genomic information and mesh it with that. So I like the idea that you do multiple things because this is one I think it was one of the most important things that the Institute could do. Yes I'm also sorry I missed this meeting because it sounds like it was a doozy. It sounds like one of the fundamental issues here is that you have to have a session where everyone agrees on a definition of function for that part of the meeting. I mean function can be anything from what is a non-coding piece of DNA do to what is a protein do to what happens when you alter a gene one way or the other and what kind of phenotype is it produced. Different people define function different ways and and there are very different ways of defining it but I think it's still perfectly not only valuable but incredibly important to have that discussion around at least temporary definitions. Everyone agree all right for this part we're gonna call function this and then go ahead then all right but now for this part we'll call function that and then have the discussion. I think having attended the June meeting on data integration that Mike and Lisa and Wiley led and having attended this other meeting it was really clear that what made the first meeting a really great success in my mind is that we had a really crisp documents going in and goals about what we wanted to talk about and what we wanted to accomplish by the end and I felt you know that it was a really productive use of everybody's time I felt like it was it was not just another meeting at NIH where you're just gonna you know who knows what's gonna come out of it right I thought it was a really important meeting and I felt that a lot of ideas got crystallized by having everybody come come together whereas I think the second one to me really opened up my eyes to a lot of the complexity and perception that different folks have and at the end I just felt more confused as I left that I was going in and and I think you know that's why I think we do need you know another meeting but not necessarily with the same group maybe with a subgroup or you know or a set of meetings that have a crisper agenda and what we hope to get out at the end right so so the two meetings are I was involved in in both data aggregation in this one and I agree with you but I think that this one had the problem that Bob just described quite well which is what what definition of function is really important here and I think that's very I think that's very hard because if you talk to a bunch of smart people you'll get a bunch of good answers so I think we can still have a very practical meeting with saying you know if the goal for example is to better annotate existing GWAS catalogs using functional data okay let's say that we want to have a particular sub meeting on that I think we would have gotten a different set of results then if the goal was how do we best use computational models to make better experiments that we can use to improve computational models and that was one of the things that to me was you know those two discussions were happening at the same time and they were kind of cross-talking and it just was unclear like okay what do we want to get out of this at the end I don't you know no more comments thanks very much