 Okay, so this session is formatted just a little bit differently. They're all a bit different to keep us on our toes. So we have 45 minutes in total and we're going to have two 10 minute presentations and we'll save all the questions for the end. The session title is future clinical implementation roadblocks and opportunities. And our first speaker is one of our co-chairs of this meeting, Judy Cho. I introduced Judy at the beginning of the meeting yesterday. So I think we'll just jump in and give Judy her full 10 minutes for her talk. Thanks, Sarah. I'd like to just start by thanking the NHGRI team and like other members of the co-organizing committee. It's been a lot of fun seeing this go from an idea to actually happening. So what I'd like to cover in the next 10 minutes is the relationship between kind of genetics, multi-omics, and therapeutic targeting. And start by talking about the classic loss of function of protectable yields where one can make the argument that in the absence of multi-omics, we're really understanding the biology at all. You can go straight from genetics to successful therapeutic targeting. And if you think about that, obviously at the top of any list, there's going to be PCSK9, which has been mentioned a couple of times. And this work from Daniel MacArthur's laboratory, as well as obviously a large consortium effort, looking at the ratios of the predicted loss of function and the expected. You can see that this obviously is the kind of the classic example of where specific targeting results in a direct therapeutic agent. Using complementary approaches, King Edel had published this a couple of years ago where it's a systematic search of does genetics help you in direct therapeutic targeting. And the second half of this red underlined area is when the causal genes are clear, the success of human genetic evidence increases approval twofold. But a particular interest is looking at the first part of that kind of red underlined area. And the genes with genetically supported targets are more likely to be successful specifically in phases two and three. And along those lines, one of the major therapeutic targets across autoimmunity is the interleukin-23R pathway. So coincident happen at the same time as the GWAS when it first started getting active in 2006 and 2007. Phase two studies of monoclonal antibodies that block both interleukin-12 and 23 were occurring. And really a striking what was evolved for the next few years was very clear is that if you look across these autoimmune diseases, those genes that have the associations that interleukin-23R, and specifically, there are multiple independent alleles with an IL-23R that are associated, but one of the more important ones is this arginine-3 and when glutamine, the major allele is arginine, the protective loss of functional allele is glutamine. And so across autoimmunity, those diseases that actually show associations to IL-23R actually correspond quite well. There's a high correlation with blockade with efficacy in terms of blocking the IL-23 pathway. So that's the good news, but there's only a handful of examples where you clearly know the gene. It's of high enough effect sizes. And so because of that, for most of the genetic associations, it's less clear you have a lot of biology that's involved. And there's one example. So of nod-2 and using nod-2 single cell transcriptomics and ultiomics and Crohn's disease. So inflammatory bowel disease affects not quite 1% of the American population. It has a peak age of onset of late childhood and early adulthood. And its clinical course is one of relapsing or remitting inflammation with progressive tissue damage. And for many, many years until quite recently, the major agent that's used to treat moderate to severe Crohn's disease and less ulcerative colitis is anti-TNF. And so in European ancestry, Crohn's disease, the largest effects risk alleles are nod-2 just associated with Crohn's about 20 years ago. It's Crohn's specific. And it tends towards a particular complication of disease of structuring complications. And so for 20 years, we've been wondering how can a loss of function alleles in a microbial sensing gene like nod-2 result in an increased risk for Crohn's disease. And we've made substantial impact on this led by Shika Nayar in the laboratory published a couple of months ago. And it was initially started by the observation of Part A, again, looking at single cell analysis of inflamed versus uninflamed tissues, as well as peripheral blood mononuclear cells on the top of A. And our initial observation was if you look at the second column, looking at nod-2 or the heat map, we can see associate the expression of nod-2, not just in the macrophages and the inflammatory macrophages, which a lot of the work of nod-2 and the intervening 20 years is focused on. But at the single cell level, we also saw expression of nod-2 and activated fibroblasts. And this kind of was, we were interested in studying then, this caused Shika to then want to study the myeloid stromal niche, because the intestine is one of the unique tissues whereby the residential macrophages are continually replenished for blood monocytes. And so you can also see in the activated fibroblast data in the red rectangle, the key genes that kind of characterize this activated fibroblast signature, PDPN notably, MMP3, Interleukin-11, CX-13, as well as WT1. And so in Part B, what Shika then did was sorted within the inflamed ilial tissue, these double positive cells designated by the purple asterisk CD14 positive, which are characteristic of blood monocytes, as well as PDGFRA, which is more characteristic of fibroblasts. And so within the biopsy single cell suspensions from inflamed versus uninflamed tissues, we can see that the activated fibroblast gene signature of IL-11, CHI-301, PDPN, and WT1 is substantially higher in inflamed versus uninflamed tissues. Again, the WT1 gene is a particularly interesting transcription factor involved fundamentally in development and transient gene expression we think plays a role in the pathogenesis of this disease. What Shika then did was work with Kyle Gettler in the lab of then taking differential gene expression between activated and non-activated fibroblasts, developing these gene lists, and then projecting them onto the pediatric and subsection cohort analyzed by, again, Subracucathezen and Greg Gibson, which is a bulk RNA-seq data set at the diagnostic endoscopy pretreatment. And you can see that there's correlation of both the activated fibroblast signature score as well as the inflammatory macrophage signature score that correlates with a number of nod to risk alleles in a dose dependent manner. And this is the key in vitro study that Shika then performed was that, again, given that blood monocytes then can then travel to the tissue and then result in altered differentiation, the chronic two week culturing of blood monocytes normally results in a mix of both fibroblasts as well as residential macrophages. But what we show that is that nod to activation via muriable dipeptide, which is a minimal bioactive component of bacterial peptidoglycan, tends to bias towards macrophages. And so fundamentally there's a defect of a failure to sense MDP or muriable dipeptide results in impaired and ultra differentiation and dysregulation of this myeloid stromal niche. And then, in addition to this, what we then did was we took the single cell data of activated fibroblasts as well as inflammatory macrophages stratified on nod to genotypes, and looked at the genes that were differentially expressed between nod to carriers and nod to non carriers, and then performed pathway analysis to identify one of the transcription factors that are regulating this gene expression. The biggest signals were included stat three WT one, but also NF Kappa B. And so NF Kappa B would be targeted by anti TNF. And so we were particularly interested that a lot of this transcription factor analysis implicated the stat three pathway, in the arms of this dysregulated fibroblast myeloid niche. And so putting it all together. She could then perform experiments with a small molecule inhibitor, known as bazidoxyphine. And this was part luck and part knowledge of the literature is that we were looking at targeting a number of biologic pathways that have been implicated in anti TNF non response in particular Fiona powers work from Oxford. And we had established that the OSM cytokine was particularly important and was elevated in those individuals that didn't respond to anti TNF therapy. Now OSM is part of this GP 130 family. Other key cytokines that are involved in this include interleukin 11 which plays a key role in fibrosis pathways interleukin six which plays a key role in growth and proliferation of the myeloid stromal niche, and that you can target all of them by targeting GP 130 which is bazidoxyphine, which signals through that stat three in this activation pathway is then blocked by by bazidoxyphine. One minute. So, putting this all together in terms of the genetics. We started with not to in Crohn's disease 20 years ago, single cell RNA seek implicates that not to is expressed not just in macrophages but also in fibroblasts bulk RNA seek we use both these inception pre treatment cohorts that the diagnostic colonoscopy, as well as this data set from artists that out, where you're having sampling pre and post anti TNF treatment to identify a new treatment target that is untargeted with present therapies. And so obviously making the connections between these these things is how we're proposing this new therapeutic target. So, Nancy Cox started this workshop by emphasizing the importance of repurposing across related traits and this is a potentially a very highly highly effective means of both plyotropy and unexpected plyotropy. Because there's only a limited number of these high effect loss of function protective alleles, we need to move towards multi ohmic integration. So, the care cells are important. We've talked a lot about blood versus tissue. And again as we try to think about therapies which is where we want to go for genetics for and be our obvious, we need to validate, and we need to de risk the next critical step in treatment development. I have a number of great collaborators both internally and externally. So be happy to take any question, or actually, we'll go to the next talk. Judy, thank you that was really elegant data that you shared with us. So let's move on to our next speaker. Yes, as Judy said please keep your questions at the ready. So I'd like to introduce Dr. David Craig. Here's the professor of translational genomics and vice chair of the Department of Translational Genomics within the Keck School of Medicine at the University of Southern California. Dr. Craig's lab is focused on developing approaches for integrating and modeling data spanning from the individual nucleic acid molecules and cells the systems level clinical measurements with the ultimate goal of improving our ability to understand disease diagnosis treatment progression and prevention. And Dr. Craig's talk is titled Perspectives and Lessons Learned from Moving Genomics into a Clinic. Take it away. Thank you guys for inviting me. I'm really quite excited to be here and I'll convince you of that in a few minutes. So I want to provide our perspective or really perspective. And I'm going to use kind of examples. They're more science based more action based and so if we go ahead to the next slide. My research is really looking like all of us at integrative translational genomics and really kind of integrating across time and size and we all see these different opportunities and we go to the next slide. It's important I think then where we kind of start and for me it starts really with a snapshot of what the NHGRI website looked like exactly 20 years ago today. And that's because when I started working it was really with individuals if you click go ahead and click from that decided that leaving leaving an environment and going to one where they had the ability to implement genomics within a clinic and maybe build their own setup was so important that they were willing to leave this home of genomics NHGRI and move to Arizona and to be really free and clear from that. And so in that sense it was the biggest priority of translation to them and they named the institute that. And so from my perspective that's where really I started. And so for the number one priority the last 20 years was getting clinical tests. That was the number one thing. And so if we go to the next slide. We see kind of a 20 year transition for me and we start in the upper left we see large scale consortium that we're involved with activities and these are really important these are things that existed like 1000 genomes that helped build a framework. We had the ability to build the clinical labs, and that allowed us probably by about 2007 eight nine to be able to look at integrated analysis Barney and DNA for prospective prospective trials, the idea that rare diseases we were really starting to see some impact Nancy talk pointed to the fact that was over only doing about 50% of the time for those. But when we started a few years before that we were only figuring those out at a 5 to 10%. So I'm going to call that a 10 fold increase in the rate of diagnosis. And in cancer that's an area I feel that is a little bit neglected. From a somatic side, we were seeing a big impact in terms of therapy treatments. And so these are things that are really became powerful and they kind of led us to a transition where we really saw that we were reaching our limits. And so with all of these activities. So the next is, I think what we see is a value of multiomics. But what the action means is, we decided a lot of us did exactly what happened before and we decided was so important we need to be within a medical system. And so we left T gen which allowed us to build clinical labs with indirect to land within USC which had an integrated healthcare system, which allowed us to see the EHR, and the key message I have here is, what I felt was important. I left the home state I lived in for 35 years, moved to LA sort of several of us. And in fact, T gen moved underneath a healthcare system, all to get access to the HR because what we realized was something else. It's not just about having a genome wide span. It's to be able to have a expansive look at the HR. We would see how information that a physician would have accessible really change perspective and it was context so go ahead and go to the next slide. And the next few will be fairly quick. So, this is my main thing and I want to say what my kind of key messages are EHR integration. I mean you got to look at the actions. All of us who believed that clinical lab tests, we're willing to walk away from that, and to start over just to get within the EHR, and this is entire institutes built around that so I think that's important. And it's really important to understand how we do that in a way that respects sovereignty and privacy. When we're talking about reaching populations that have sovereignty in our part of America, I mean we've got to understand that they need to. We have to think about what privacy means to them, yet we still need to mandate some level of data sharing standards matter. I think the one thing I would point out is is that NHGRI did not necessarily push in some of the somatic landscape. And I felt that they could have been some of the impact that didn't happen is, you know, we don't have a great standard like in a one trade seven eight VCF and bands are kind of germline specific. I want to emphasize about how that's done and sometimes it requires private public partnerships. And I'll talk about that in a second. And then I want to also mention that multiomics is impacting the clinic. And I'll highlight that as well so go ahead and go to the next slide. And so open data standards impact, go ahead and just click. And if you just keep clicking through this is where we had put together an effort to really build out and understand colo eight to nine as our standard. And we had to do it kind of on our own. And it was partly because this is something that was not a priority for really anyone at NIH and allowed us to look at reproducibility that actually in 2020 we discovered it wasn't always the perfect reproducibility. We could show how we kept getting different phones the same way single cell opened up all sorts of insights go ahead click to the next slide. Those were relevant for moving things into the clinic one thing people talk about is a lack of actual data to train students and I'm going to tell you if you want to do one thing. You actually create disease specific resources like how to see I'm sorry like how 1000 genomes is where everybody can access it can be accessed in a class to learn how to study disease. How to how to identify genetic basis of disease from a genome. Find me a genome where we can do that. We all go to 1000 genomes healthy people are tumor normal cancer. We're not training students because you have to join a lab to see data in that by that time it's too late. Next slide. So I want to emphasize a few real important private public partnerships and PD is something we're a part of it and we've released there's a paper coming out 10,000 participants really rich phenotyping. I think about 9,000 whole blood RNA samples. Most of it are at four different time points incredible data that goes about a billion reads overall 200 million at month zero 12 or and so forth. It's missed sometimes because it's not part of a traditional framework it's probably amp TV. If you want to see some really incredible data that's fully available for everyone. Take a look at that. Next slide. Because it really does tell us some things everybody keeps talking about needing aging data so this is tax change five time points and just walk through these go ahead and click. So these are some of the samples. There's a thank you diagram showing the flow. We have we have whole genome sequencing. You can look at LDL and the healthy controls. This is really great resource and I guess I'm right here just promoting it. And so if you go to amp PD or you can learn about it go ahead and click and right here. This is about some pictures from a paper coming out a few weeks in nature aging, where you actually show the progression what we're seeing is enrichment of neutrophils over our study. This is the same Sam the same individual at four different find points. So we're doing exactly what people see. And what we see is really a significant and change in the neutrophils. And that's where we're looking at whole blood again and then we're thinking ourselves boy single cell would have been nice. But still, this is a tremendous resource and there's a paper coming out. Next slide. So, oh, we took the same 90 same we took 96 individuals and we did a series of multiomics and part of another consortium called found in PD. This isn't coming out soon to go ahead and click twice. And these are some of the individuals involved. And if you want to take a look at it we have single cell RNA seek from IPSC's attack seek. Hi see just about every only so it's the same individuals, but we generated IPSC's and neuronally differentiated them and made this as kind of a resource to see what parts of this are useful so it's a great resource found in PD. Go ahead and click. We're going to get the other aspect that we're trying to integrate in a spatial and to understand spatial. That is something that's we're finding new areas of if you go ahead and click twice. We're finding there are new ways to think about this one of the things we found is, you know, RNA helps us but we were looking at the DNA information we were finding spatially distinct click once. And that showed loss of heteroesicosity. And that was probably because there was a EGR application taking off in one tiny corner that from the h&e we can see, we could see some specific things happening that the pathologist was excited go ahead and click the next slide. And we started to ask these questions. This is just a slide to remind us that hence history does matter when we made clinical tests. First clinical tests really did have a bias false positive rate go ahead and click twice where depending on your underlying population your number of private mutations led to false positives. Go ahead and click next slide. And I just want to say that we're going to be impacting clinical management already. I know about clinical tests and I know the impact and I also know that we filter away a lot and what we filter is based on databases. The genome database started impacting things so early and I'm already saying go ahead and click on the next few slides to take us to the end. So computational aspects of ACMG are starting to allow us to add in new filters to expand beyond coding exons to three, four bases within the exons to find cryptic splice site variation. And I can see how that filtering scheme will change and this is an example where we find an entronic variant that a 20 year diagnostic odyssey gets solved. I'm really interested in this common variant aspect of the other 50%, but beyond the exome. That's important to and so that pretty much brings me to the end and you go ahead and click through where the clinical tests kind of takes us. So these are my key messages that I want to end with. So go ahead and in. And by the way, I am at Disneyland right now. I found a corner of Disneyland, but apparently I can't protect my slides so thank you for being at accommodating. I could actually mask my background that's what's great about these pandemics, and you guys can get a live shot of Disneyland. Oh, thank you so much, David and thanks for joining our workshop while you're on vacation. Wonderful. So that was, that was a phenomenal talk as well and I know it's so hard to pack in so much information into 10 minutes but you both did a fantastic job so we have 20 minutes for our discussion, just a reminder that the session title future presentation roadblocks and opportunities and we've heard you know Nancy kicked us off with her talk in the beginning, you know, we looked at the portfolio analysis we realized there's a significant and growing investment in omics research at the NIH but there was the question of we haven't really seen the ability to move much of those that biomarkers into the clinic so we want to we want to hear from all of you on what we need to do research and opportunities. Other other things that we need to make this happen so either want to hear about that in general or specific questions for our speakers based on their presentations. And I just wanted to emphasize that the comments Nancy made the idea of Federation is one that's really intriguing the undiagnosed disease network kind of touched on that that is a new area of data sharing. It's something to consider. Thanks David. We have any any questions any one want to weigh in I mean we we certainly heard throughout the last day and a half. I think there's a lot of comments about the need for the validation of these biomarkers I think I'd have a question to back to put back at the group is there. Is there a role for NHGRI or the genomics community here to develop sort of standard validation pipelines or is this really so specific to the particular disease being studied. Yeah, the one thing I would say that was most important. I was at 1000 genomes, and I sat in front of Daniel MacArthur and he was talking with I believe. Oh Christo and they were just talking about trying to create a database that became genome. And what happened the first implementations utilized public databases to do filtering. Before validation occurs, bioinformatics methods are utilizing multiomic approaches. So, you know, when genome came out are the exact database 1000 genomes came out. It was part of the valid it was part of the filtering process so in other words, it, it doesn't actually have to be clinically validated to influence what the final result looks like what variants show up. The minor allele frequencies from 1000 genomes shaped every single clinical report going forward. Thanks David. Yeah, that's a good reminder. So we have. I think I saw Greg's hand up Greg Gibson. Yeah, thanks Aaron I was just going to say that nobody's really talked about a CMG guidelines yet. And so in the context of multiomic for example using RNA seek to identify a splice site variant. What, what challenges of those have been implementing multiomics in the clinic come up with in terms of establishing pathogenicity and other guidelines need to be revised. And that's really difficult so in that one case, I, we go, we went a lot into it so you know, there's so many different ways to look at RNA seek and to validate it. I mean we built an assay. That was a technical approach for actually putting everything in but at the end of the day, we found it was easiest to, you know, use the evidence with the pathologist under certain guidelines and would only get really a V us on the front page and then usually what would happen is that would be enough for the clinician then to debate the evidence RNA is really a difficult clinical biomarker to implement. And I guess I felt that it's more important to start thinking about how to integrate ph our data because I think we need to go further. We did it we tried. And ask a different question we did X and activation. It's a tough one. Put five years into it. Thanks. Thanks, David in the launch in. Yes. Yeah, that was excellent sessions, you know the whole day was amazing. One, like a general comment is that, you know, especially from a lot of talk about risk prediction, how much data can improve risk prediction. And I think in that context, other than, you know, validation, I think we need to define what is clinical utility. In particular, for example, for PRS, polygenic risk scores and I had a recent reporting standard paper, you know, with the with the committee and where we kind of wrote down what is the clinical utility how it should be reported. And I think we need to do similar things for other biomarkers as well, just showing that it has some predictive value. And it doesn't mean that it is going to be clinically useful. So we have to, you know, come up with criteria as far where in a particular disease. How do you show clinical utility. If you're, for example, thinking about, you know, that it will help you to read certification or developing who should get certain therapies. So that that's something we need to invest more time. I think that's really a big challenge. And it looks like we've got some support in the chat from that as well. So, I think the long and that's a that's a really good point and that might be something that comes back up in the recommendations as well I see a bunch of other hands so let's, I think I saw to Lee then Ali. Yeah, I wanted to get back to a point that I think Judy mentioned a little bit and Craig and others of, of kind of talking about mechanisms, getting to disease mechanisms where I think that NSGRI is doing a huge amount of super valuable work setting of things like IGVF and other studies like really getting into how does the genome work, which I think is that the core of NSGRI's mission. And I think that thinking about multiomics in more sort of like a, like a cohort setting has actually a really important role here. I don't think that we actually know how do we stitch together information that comes from Chris bring some cancer cell line and multiomics and naturally current genetic variation that we observe in human cohorts, and both can be informative of mechanism, but sort of thinking about principle ways of how to put these things together is I think something where I think NSGRI could play an important role. I don't know if others have comment on this or. Thank you. Thank you to Lee. That's a good point. So Ali and then Mike. Yeah, just a couple of quick comments. You know, I want, I want to emphasize the data access. You know points have been brought up a couple of times now access to EHR data. I mean I think that's supremely critical for, you know, validating some some of these prediction models. And we have, for example, the top med data sets out there, they published a paper on harmonizing phenotypes, but they have yet to distribute the code and and phenotypes as related to that gets a little little frustrating. Over time, I think is a big problem. The, a comment on, you know, Greg's on the RNA RNA seek and maybe it relates to some of the end of one sort of conversations that were had as well. Julie knows as well that there are some methods out there, taking population data, like a page one who worked with Pigeon harmony who worked with Julie is developed one of the methods where you where you can take, you know, population data to define what the normal normal range is or expression of a gene or, or, or maybe some other biomarker that you're following in a end of one study and sort of establish the reference range based on the population and look at whether or not it deviates from that reference range. I think we have a lot of good comments also on personalized reference ranges as well. And then finally, on the PR standards for clinical utility I think David Craig was about to say it's very challenging I think for you know it's easy it's easy to put standards for risk stratification and statistical measures but once you get into clinical utility I think it becomes pretty indication specific so coming up with a with a particular standard for clinical utility that works across the board I think might be a bit of a challenge. I have a question for other folks. When Nancy talked about federated access to EHR. I've always been impressed with what the undiagnosed disease network now I'm just wondering if it's just me and a few others who think that that's a potential interesting way to also worry about when she talked about getting attacked by cyber attacks at the same time. Is there anyone else who feels that federated EHR databases like how undiagnosed disease network would allow querying at different sites for different variants, you know, expanding on that's worthwhile. Just anyone weigh in on that Judy the feasibility standpoint. I'm not sure exactly how that would work operationally as a point of principle though with respect to electronic health record. The patient is the owner of their data right. And so with the New York state and New York City, you know you can basically sign off that yes go ahead and share my EHR with these folks. And that that may not scale efficiently so I'm having trouble understanding a federated system across health systems. Patient directed can definitely work but again that. Well, so what undiagnosed what the undiagnosed disease network did with Broad and others in they used an off zero to connection framework where they got the ability to query variants across different sites, and to share a certain information under that protocol in you know I guess they kind of stumbled in that area and that's where. Yeah, so I'm concerned about the technical aspects being too prohibitive but it's kind of if we're going to talk about lofty goals. I mean, if we were in psychiatry or psychology we would not be where we would be very impressed with how genomic shares data. How do we share things. Is it just easier to like you say get approval to download it and put it together somewhere else on a safe site. David thank you let's move on to Mike and then Nancy, and there's a really good conversation in the chat if you want to take a look at that, David, go ahead. I retract the use of the term own. All right, Mike and Nancy. Yeah, just stepping back. Now it's true I miss some of the talks because of other commitments but it seems to me. Moteomics is super powerful for discovery for analyzing systems, you know for virtually all diseases right and every own contributes, and that's been true for everything we've done with it's pretty a metabolism. I learned some pretty important stuff and can lead to actionable information. I think one bottleneck is when you just come up several times getting into turning things into clinical tests and I think a lot of valuable clinical tests these days are going to be multi analyte tests. And, you know, clinically we're used to one off test, you know, or maybe a few analytes but the reality is I think the the test of the future might be analyzing several hundred things all with absolute quantification so that you can tell the results from one to the next. Now quantification is relative quantifications there but absolute quantification is usually missing in most Moteomics assays so we definitely need that. But I think that is a bottleneck and I can tell you for everything we've done oh we have a new marker for insulin resistance we can replace the $6,000 test with, you know, basically 20 analytes but that's a real pain in the ass to develop that. And I think everyone probably on this call has some sort of similar situation for their favorite disease situation where again that they can go after the major one with maybe one or two analytes but if you really want to get a more comprehensive picture of disease and you know most diseases are really syndromes to involve a series of different underlying defects are all bundled together diabetes is a great case of that. And so I had think having multiple analytes to be able to tear that apart. But if we can set up ways of being more efficient of that I think we have accelerate clinical impact and then of course you have to show utility once you build those assays. Thank you Mike. So Nancy, I'll turn it over to you and then if we have time I want to come back and also talk about sort of our discussions we've woven in about generalizability and promoting equity, as we're thinking about moving into the clinic but go ahead with your question. So I just say that I think it is it will be possible it is possible to federate important kinds of queries across many health care centers now. It's just a matter of coming together and agreeing to create some some very straightforward sort of variant by diagnosis kind of pre specified tables that are lookups that that allow, you know, a query to be fair, quick, you know, put out across a large network anybody that has the Cape. I mean this is a lookup not a not even a calculation it should in many cases would be pre established as a lookup and at least as a start, there are very simple ways to set up federated queries across large systems and and what David described is absolutely possible to do in many health care centers, even now. As more and more genetics data come in, those will become more and more powerful queries and that's why it's so valuable to think about trying to establish the right governance and get get a lot of buy in from as many places as possible to enable these these things to start moving forward because as the data come in this becomes a richer and richer resource that that accelerates things faster and faster in a good way for for helping our patients and for helping us learn better than how to help our patients even more. And I absolutely agree with Mike, we have to move towards these multi analyte very straightforward to things that are very straightforward to to validate enabled to be validated. But boy, I also think we've got to push more for things like RNA seek that are so rich, even though it's hard, there has to be a way to to pull information from from some of the rich omics that that can be useful and so that we have to push on both fronts make that do the take the easy wins where we can and keep pushing on the harder omics, because they are so rich on the on the RNA, on the RNA question, we can get to very straightforward a kind of a technical RNA, where it's basically all these measurements are made in a plea environment, and then not make an interpretive report is that of value, because that is very doable. Does anyone want to comment on that question. Yeah, that clear line is a big line. You're either clear or you're not you can either return the results back or you can't. That's my experience. But other people may have different experiences. It's a it's a big line to cross. That's just my experience. In our in our closing three minutes here again just coming back to our big question so we're, we're hearing a little bit about some of the steps that we need to, we heard a lot about the barriers, some steps that we need to take I mean we're hearing about the EHR integration. There's been such rich conversations about the data sharing. What about thinking about generalizability and health equity in two minutes remaining we don't have much time but does anyone have anything they want to put on the table here need to do more. There's a fundamental challenge where major medical centers tend to have a bias towards a very not representative kind of view of the United States and I mean, you have to fight to get to the places that are more they don't patients you don't end up there they end up with later stage cancer different cancer right there we end up at county hospital systems and we've got to find ways to engage them on terms that they understand a Native American population has a very sovereignty. Can they have some say over their data that's a little bit distinct. I think we have to be flexible sometimes in order to gauge engage certain groups and we it's worth it. But there's a real trust challenge. There is a trust challenge and recruiting. And we have to really address that through workforce through many things. Thanks Judy, I'm going to give to Lee the last question or comment. I just wanted to comment on what was said and elaborate a little bit on the do more aspect so so now that there's a lot of investment in making. She was data sets more diverse and this is moving we have to make the functional interpretation of references more diverse as well, there is no cheats resource that actually covers diverse ancestors. And as long as we don't have those we may end up in a situation where it's like okay so now you did the she was this but we're super far behind when it comes to functional interpretation. So yeah, those gaps just need to be need to be filled. That is a great important point to close on so I think we're going to wrap up this session. Now there I mean I can't there's just such an amazing conversation happening on many levels in the chat so we're glad we'll be able to capture all of that and we will turn over. And the remainder of the time to Howard and Judy for recommendations.