 Welcome back from lunch everyone we have a couple of concepts to present to the council and I just want to take a moment to explain to the new council members that before any Institute in NIH Can publish a funding opportunity announcement that has a set aside of funds associated with it It must be approved in an open venue by an external advisory group now We always use the council to do our concepts so that you guys know everything that's going through The system here. So first let me also remind you that There will be a formal vote taken at the end of the Terry's going to give a presentation We anticipate questions in a discussion at the end of that I will take a vote asking for you to approve a concept. So Terry you want to Emerge comprehensive genomics risk assessment and management. Sure. Thanks Rudy And I'm I'm pleased to have the opportunity to present our proposal for the next phase of the electronic medical records and genomics or Emerge network. This is based on input from our advisors scientific planning meeting We held specific to emerge and the current state of the science So we are Proposing to use genomics information in the context of other clinical information for genomic risk risk assessment and Okay, so this doesn't really want to work terribly well So that may be a problem. Well, all right So for risk assessment and management of multiple common complex diseases Recognizing that in order to do this in a scalable way with busy clinicians that it has to be integrated Pre-seamlessly with the electronic medical record and electronic clinical decision support So it seemed like a program like emerge that's focused on electronic medical records was kind of a good place to do this so how we have a number of aims four of them are our sort of Right two of them are retrospective and kind of methods development and then two are more prospective and analytical We're proposing to use by our repositories linked to electronic medical records to develop Implement and disseminate genomic risk assessment and management tools for clinical use First we would calculate validated polygenic risk scores for several complex diseases retrospectively Using EMR defined phenotypes and available data sets to the most for the most degree These PRS's have not been used on electronic phenotypes We expect that they will validate but we don't know that so we sort of need to prove it and it might be you know for For some of them it does and some of them not so well And then share those distributions associations and other characteristics out widely And then develop the EMR based methods to communicate genomic risk profiles and relevant clinical recommendations Based on polygenic risk information as well as family history and other clinical data and Then prospectively recruit in genotype 20,000 individuals of diverse ancestry Prospectively calculate their genomic based risk of for selected electronic phenotypes and then provide risk estimates and management recommendations Management being what the physician and the patient should do about those risks to reduce them To them and their providers through the EMR or patient portals And then use EMR based methods to assess provider and patient uptake of risk recommendations and the impact on related clinical outcomes Emerge would also continue its efforts to use biorepositories linked with EMR Search by expanding and enhancing electronic phenotyping identifying genomic associations with e-phenotypes developing and disseminating ECDS tools enabling integration of genomic findings into the EMRs for clinical research and care and then disseminating those methods tools and best practices to the scientific community To start with a few definitions by genomic risk assessment we don't mean only use of Sequence variant alleles, but it would be calculation of risk for complex diseases from not only risk allele information Other genomic information that might be current at the at the time that this is done such as transcriptomics for example Or or epigenetics a family history and clinical information by polygenic risk scores right now We mean just risk assessment based on risk allele frequencies. Those are primarily identified through genome-wide association studies Family history would be using is typically using patient-driven web-based tools that integrate into the EMR such as ignites Me tree that NHGRI has has had a significant investment in but could be others or clinical information That's non-genomic information such as lab values anthropometrics past medical history personal habits like smoking Etc Risk management would include an implementation of further testing for example such as mammography if indicated bone density testing specific gene panels even or interventions such as drug treatment surgical Interventions et cetera to identify treat or prevent early disease and clinical risk estimation would be non-genomic information As I mentioned before such as medical history physical findings and laboratory values that up until now Physicians and patients have been basing their risk estimates on A diverse ancestry would be defined as non-european plus Hispanic Latino ethnicity per OMB directive 15 that that's these categories here These are self-identified and underserved populations are defined by the health resources and services Administration and the National Institute of Minority Health and Health Disparities as areas of populations designated by her says having too few primary care providers high infant mortality high poverty or a high elderly population And there's a website and tools to be able to find these places There's a map that shows them all you can see some in the middle of the country here But also in in various areas around the country Just to remind you emerges has had three previous phases It's currently in its third phase the very first one began in 2007 Can't when we ask the question can electronic medical records and biobanks be used for genomic research the answer Was resoundingly yes, so the program continued and we at that time We're focusing on genome wide genotyping electronic phenotyping and GWAS studies in our second phase We wanted to know can genomic findings be applied in clinical care And if so how so some implementation pilots were begun We expanded into pediatrics and pharmacogenetics and then continued the e phenotyping and GWAS studies And then in the current phase can sequence data and clinically relevant genes be used to assess penetrance and improve clinical care And there we added some sequencing also more mature clinical implementation e phenotyping And throughout this we've conducted LC research into each of these topics So is this are we ready for this now well recent advances that make the study of genomic risk assessment both manageable and Feasible is our clinically certified dense SNP arrays and imputation algorithms are now widely available And are relatively inexpensive comparatively Consensus approaches to interpretation of actionable variants have been developed by groups such as ClinGen the American College of Medical Genetics and Genomics and others Professional guidelines for clinical use of actionable variants in high-risk individuals have been promulgated and there are further Efforts to develop such such guidelines among various professional societies Polygenic risk scores and just in the past few years have been developed for multiple conditions Including atrial fibrillation diabetes. I'll probably both type one and type two ADHD coronary disease Alzheimer's breast cancer other cancers Automated tools for systematic patient driven collection of family history are now widely available and integrate with the medical record and Pharmacogenetic variants that predict altered response to commonly used drugs have been not Identified and are known to be carried by died just about all of us But there are a number of gaps that need to be filled before clinical adoption of genomically defined risk assessment and management can really become Widely adopted these include development and validation of EMR tools for seamless integration of genomic risk estimates into the medical record and Delivery of recommended clinical actions in a user-friendly friendly manner using ECDS predictable of EMR derived phenotypes from Polygenic risk scores has to be assessed Estimation and validation of PRS in non-European ancestry populations is a major gap This is an area that's being worked on Validation of PRS based on e-phenotypes types also in non-EA populations needs to be done Assessing and uptake assessing the uptake of risk reduction recommendations across a range of conditions also needs to be done And estimation of achieve achievable changes in related clinical outcomes So so how much does this affect you know LDL cholesterol lowering for example or increases in bone density or that sort of thing We really don't have that those kinds of things quantified just yet I think that the proposal we have for a merge can address all of these things except the estimation and validation of PRS and not European ancestry populations There are a number of other programs that are doing this currently and are expected to do it in the future Our genome sequencing program where a number of the polygenic risk score data have come from to date The million veteran program the all of us program to the degree that their data are are available The page program our population architecture and using genetics and epidemiology is is an entirely minority population The top med program of the National Heartland blood Institute that is again Extremely diverse. So we're hoping that these kinds of programs will will come up with a fair amount of information On using polygenic risk scores and validating them in non-European populations So with the objective of retrospective validation and adaptation of the PRS to EMR defined phenotypes We would use publicly available densely imputed SNP data in 83,000 currently available 83,000 emerge participants and actually this number is expected to go up to about a hundred thousand Shortly and then other studies to calculate Polygenic risk scores for say and again This is just an estimate say 20 complex diseases and then identify appropriate thresholds Which could be would be defined by the investigators and could be say the top one or two percent of risk or people who are at a three Or four-fold increased odds of disease, but whatever risk threshold one sets You would need to be high enough that the that people are truly identified as being at high-risk and distinguished from other groups For risk reduction recommendations based on current guidelines and budgetary constraints because there are only so many people in whom we can Intervene and then determine distributions of risk scores across key demographic subgroups such as those defined by Ancestry or by geography or age or or other things and then modify the thresholds as appropriate for differing allele frequencies or clinical characteristics This just gives an example of some of the risk estimation that does differ say from between Finnish men and training him Men as shown here by Abraham Abraham at all Note that these risk scores do nicely separate out the finish men maybe not quite so well Although I upper at older years it does well in framing here But mainly because they're different baseline risks You're seeing the differences there and similarly these data from desiccant on Alzheimer disease survival free of Alzheimer disease does a very nice job of picking out the highest risk group That ends up without time or some fairly at fairly high rates You compare risk predictions with specific electronic phenotypes Modify the thresholds or the the phenotypes be used as appropriate to get a good concurrence between them and select some subset You know, let's say for the sake of argument 15 risk algorithms for prospective Implementation and then estimate as feasible the added value to the degree that we can the added value of genomic information to risk estimates based on Clinical or non-genomic information. This is only one of many analyses that could be done One could look at the low risk people and there are a number of things that could be done with this data set For prospective risk assessment We would then propose to recruit about 20,000 and again These are estimates of diverse ancestry to undergo clinically certified dense SNP genotyping family history assessment and clinical evaluation risk Provide guideline based estimates and clinical recommendations to providers and patients through their electronic medical records Arrange for follow-up testing and risk management of patients who exceed agreed-upon thresholds And this is one reason to keep it at a very high level of risk because again, this is a fairly intensive enterprise Quantify the uptake of risk management recommendations So how many people actually got their tests done? How many treatments were initiated and iterate the approach is needed to reach 50% uptake? Again, this is a proposal in an estimate why 50% well as a clinician I'd be thrilled if half my patients did the things I asked them to do so so it seems like 50% was a reasonable way But one could one could argue that point and then disseminate and analyze what will be a very rich data set And you know one analysis that could be done for example would be estimate You know, what does the genomic information add without doing a clinical trial? We don't feel that we're in a stage where we were we're actually ready to do that sort of thing But one could figure that out, you know based on some of the other information in the data set Risk management would include guideline based interventions tailored to individual patients estimated risk and then the risk met Management recommendations would be delivered to patients who exceed the agreed upon levels of risk others will get their information on the risk But but really the guideline information would be given to those above a certain threshold and then design and components of the risk Risk assessment including the guideline based risk reduction recommendations would be proposed by each applicant as they apply but Have received strong recommendations from our advisors that that emerged in its next phase needs to go forward with one protocol rather than Six or eight or nine So there would be a single network wide protocol for risk assessment and management and outcome ascertainment that would be finalized by steering committee consensus Recognizing that these six words and entail a host of conference calls and meetings and other things But we we do believe that we can come to consensus on it And then risk assessment and management would be updated as needed as new and new risk information accrues So these wouldn't necessarily be static either risk assessments or risk management recommendations because this is a very rapidly changing field We'd have to find a way to incorporate those changes Data collection would include standard clinical risk factor information for a variety of diseases I just listed a few here hypertension obesity health behaviors That would be extracted from the electronic medical record at the beginning and throughout and then outcomes would include disease surveillance or screening Interventions consistent with guidelines a drug selection and dosing again consistent with guidelines Improvement in modifiable risk factors after guideline directed care other guideline directed care such as drug treatment surveillance practices other health behaviors And certainly would want to look at the cost and utilization of health care in the high-risk group For sample size estimation if we had 20,000 patients and again, this is just a proposal We may end up with a different number But if there were 20,000 patients who were screened for genomic risk using a variety of modalities About 5,000 or 25% of patients would be at the top 2% of risk for 15 For one or more of 15 complex diseases the numerate among you will probably say well g 2% times 15 should be 330% and that's times this would be 6,000 but recognizing that people can be at risk for more than one There would be 400 who would be at the top 2% for any single disease But again, those numbers can overlap so that even though for one disease. It's only 2% for 10 diseases about 18% And 15 diseases about 25% This would be primarily a descriptive exploratory and methods development program So rather than doing hypothesis testing we would instead try to describe what actually happened And one way to do that is to place a 95% confidence interval around the uptake proportion To see you know did this actually have an impact So for the proportion of 50% for each disease for 400 persons at high risk of one disease and 95% confidence interval would be about 5% plus or minus for some subgroup Say it's you know one ethnicity or whatever if they comprised 20% of the high-risk subgroup the 95% in the confidence interval would understand we do wider about plus or minus 11% And then for selecting clinical sites estimating 6 to 10 clinical sites in a coordinating center The a subcontract for genotyping would be included at the coordinating center So not separate centers for sequencing and genotyping as we've done in some of our other programs There would be two clinical site Funding opportunity announcements one each for more than 35% or more than 75% Diverse ancestry or underserved populations. This is an approach that we've taken in in each of our two Most recent program renewals. It's been quite successful for us The reason we've had to is that we recognize there are some areas that simply can't get up to the very high levels At the 75% range, but we can sort of pull them along and bring them into the 35% range Plus there are some that we can actually stimulate to go a little higher and strive for that that higher level which given the vast under representation of non-European ancestry populations in our studies and in every one studies seems like an appropriate thing to do each clinical site should be able to recruit and Followed about 2,500 patients assuming, you know Eight centers who have not previously received genomic information So these would not be people who had been in other programs whether clinical or or research And then the site should also be able to implement electronic phenotyping ECDS and outcome assessment in a comprehensive electronic medical record They should be able to provide valid e-phenotypes for a variety of outcomes to be determined by the steering committee And then there are sort of the standard things, you know, how collaborative are they? How what's their record and data sharing and productivity next brand? What's their expertise, etc? Budget assumptions were clear SNP genotyping at $350 we realize this is a high-ish price for SNP genotyping But to get it clear these are the prices we were quoted we expect that those will come down We also expect that the the costs and prices of other forms of genomic assessment will also come down The coordinating center with some cut subcontract for the genotyping There would be eight clinical sites recruiting 2,500 patients each again These are on average the cost components could be divided then into sort of fixed with our which are independent of sample size And variable that are directly proportional to the sample size and each high-risk patient We're assuming would incur an additional amount of testing. We just guessed at you know mate Let's say it's about 1,250 in follow-up for some it's going to be three or four thousand dollars if they have to get a Fancy genome panel that's only available from one offer Or you know could be less than that and again making us making an assumption that about you know One year thirty percent of that would be paid by insurance Maybe just assume about a thousand dollars per high-risk patient would be borne by the grant and then genotyping costs would be concentrated in years two and three and This is then what the budget looks like for the fixed costs for the investigators the informatics systems Etc variable costs for recruiting the cohorts and genetic and other testing indirect costs about split evenly among all of Those things for a total of sixty five million with the highest costs in years two and three And then breaking those down by clinical sites This is just again an estimate, but we would estimate about four point two million direct cost Perhaps six point six million a total cost per clinical site for eight clinical sites Fifty three and then the coordinating center with the with the genotyping twelve million So these would be the estimates that we go into RFAs So I think with that I'm finished. Oh just to give you a timeline We are here currently in mid-February Emerged currently goes until about June of 2020 and you're seeing this renewal concept that RFA would hopefully be released sometime in say June review Initial primary review in the fall and bring it back to you in February a year from now The awards hopefully in the summer and risk algorithm development and exploration would begin at that point the very earliest we'd be able to begin a recruitment. I think this is quite Optimistic, but the earliest would be April of 2021 or more than two years from now So a fair amount can change between now and then I think with that I'll stop and be happy Oh, and I should note that our council discusses are doctors de Verca Haynes and Juan and I'll start with dr. De Verca Dr. De Verca if you had comments, you'd like to make you don't have to but if you'd like to Just a couple of comments particularly about the prospective part of this Is there any had there been thought about having a comparison group? Prospectively collected comparison because I really was struck by the importance of what is the incremental benefit of? Polygenic risk scores over traditional phenotypic measures And so I think that is an important question And I wasn't clear if you were saying that you didn't feel like we were ready to actually do that So could you just expand on that a little bit more right? So so the comparison groups could be many in this in this whole large group So one one logical comparison group might be the people who are just not quite at that threshold But but close to it or it could be the entire group. That's the way a lot of the polygenic risk scores have been calculated We didn't feel that we were were quite ready for a randomized trial Which is really what you would need in order to be able to assess the question that you were asking but we think there will be a fair amount of Descriptive information that will allow us to at least get at it And I guess the other comment was you're you know saying that you won't be able to Interpret the or validate the polygenic risk scores in minority populations Underserved populations is that right? Is that you're not like other data sources to do that? I would suggest that a merge would not be the best place to develop those scores I think we could validate them, but but you know Developing them you need very large numbers with you know with dense sequencing information Which is is being generated in our other programs and programs of other institutions So so is there a thought that the risk management recommendations might be any different in underserved populations then for those of European ancestry if we feel like we're ready and these are going to be predominantly groups that are 35 to 75 percent underserved I Guess you know just for my personal opinion. No, I don't think there will be any different I really think when one is looking at genomic risk one's looking at risk allele frequencies to date the Where polygenic risk scores have not predicted as well in non-European? Groups is because they either don't have the alleles or they don't have them as frequently So the risk estimates tend to be poorer. They do not tend to be opposite One would worry about them being opposite, but but to date they have not been and I think in two years We would know much more about that from other other stuff Thank you. Great. Thank you dr. Haynes So I want to follow up a little bit on the on the diversity angle because I think we already know that the Polygenic risk scores are not going to work as well in any of the diverse populations And we don't know how well they they will work that is an area that as you say are is very much under You know development and trying to understand that so could you Think of little or expand a little bit on the timing of the recruitment of the of the diverse the diverse samples Because if it turns out that the polygenic risk scores from what we know aren't really very good for the diverse samples And we don't know what the the good ones are Necessarily then is it appropriate to be focusing on recruiting all those diverse samples at the beginning Yeah, this is a tough one and obviously we've been struggling with You'd kind of hate to to wait to recruit them as sort of you know You guys have to wait until we have the answers that we want In addition, I think you know one of the big challenges in in this field in any field is that there are always going to be Fewer data and in non-European ancestry populations always you know the data that the follow-up is going to be less The data will be less less complete etc. And so you know do we just promulgate that by saying okay? We're not going to even try we'll wait for somebody else to do it Or do we say we're going to do the best that we can and recognizing that risk scores do separate people at risk? They may not identify folks who are you know between the 20th and the 40th percentile terribly well But I think one reason to focus on the very tippy top of the risk is exactly that that you you would be more confident At least that you're getting the highest risk people and remember it's not just polygenic risk score It's also family history and other clinical information you get the people who are at the highest risk within their demographic subgroup But we we'd need to leave it to the investigators to determine how they would want to do those in an ethnic specific way Yeah, I guess I'm just a little concerned that you're recruiting people and you may not be able to to You know deliver what you're sort of promising promising to them So I think it's something you really you really sort of need to think about the other the other point that I'd like to to Talk about a little bit is the the inclusion of the non-genetic information as well because I think it would be really really useful And we've certainly seen this in some of the other diseases that I work on that the you know The polygenic risk score actually does no better than using the non-genetic information and when you add the two together You don't get any benefit from you know either one. They're sort of replicating each other So I think it's really useful to to bring in the non the non the non-genetic information I think it would be a nice focus of this to figure out what how you get that information out of the EHR I know that you know, they've been working on that for a long time But there's still a lot of variability in what those what those algorithms can pull So I think there's that's an area that would be a nice a nice focus agreed And I think you might agree that emerge would be in a good position to be able to do that So I think we can't exclude the non-genomic information I mean we're trying to do something that that will be clinically useful and clinicians are not going to ignore the cholesterol level or Ignore there, you know, there are three ants with with breast cancer or whatever But but I think trying to you know meld those together and use that information together Is probably the best way forward can can we really say that it was the genomics that made the difference? That will be tough on the other hand There there does seem to be some evidence that perhaps genomic exceptionalism works in our favor here because if you tell people that you know We've identified you to have genomic or genetic risk for something sometimes That's a little bit more Activating than just telling them they're at clinical risk and there are some small studies that have shown that So we need to that's something else we can test Yeah, no And it may be it may well be that at the very tippy top or the very bottom of the of the risks that the genetics Actually does make a difference and that's we we don't know And I think that's something that can we can find out great So we had three discussants and then and then we do okay, so so one one more and then you then you can talk Great, thanks. I actually had I guess four comments I'm actually more optimistic that the Polygenic risk scores for diverse populations will be in place to validate by the time you start I think there is a huge Pressure certainly I was talked a lot about at the strategic planning meeting and certainly Many people in the field are aware of this problem. So I'm optimistic about that I also think although we didn't talk about a lot in the discussion the relationship with the electronic health record and The goal of how to disclose this back to physicians the electronic health record. I think is critical Most physicians that's where they get their information about a patient and whenever we do research studies where they get it other ways You're not really mimicking Real life and I I hope that if anything will learn even how to do it in such a way that other Hospitals are willing to adopt it because I think we solve the problem that the big centers do I Am concerned about this issue of if you don't have a control That you're really going to have difficulty with the data And so whether these sites will just have to identify an equivalent number of patients or something or the practices The standard practice of those physicians Because I just think it will be difficult but what I was most confused and this probably relates to what Rudy said about the The concepts only being three pages long I did not understand that the RFA or the the grant mechanism would actually pay for the follow-up care At least again in the cancer sphere That's expensive if you put someone at twice the risk of breast cancer, which many of these PRS's do Then the recommendation is breast MRI and each breast MRI It is quite expensive If you shift people to getting colonoscopy at the age of 40 instead of 50 You know each colonoscopy several thousand dollars So my assumption was it was still going to be up to the patient's health care to provide that I think that if you're doing a mixed model where you pay for some of it But there if their insurance pays that actually is very difficult to implement Um, because either insurance wants to know either they're paying or you're paying because if you're willing to pay for one person Then they want you to pay for their patient too Um, so I would think pretty carefully about that part of the design No, that's that's an excellent point. I think we are hoping hoping hoping That much of it will be paid particularly the high cost things will be paid by insurance And I think increasingly that is happening We thought we needed to include some amount of money for for some testing that might not be covered by insurance So that's why we came up with this estimate But I think you're right that we would need you know among the investigators and their individual systems or whoever their insurers are There would need to be some consistency as to what would be paid for by insurance and that's going to change over time So so this is just an estimate Okay So I just wanted to make a quick comment about the portability of polygenic risk scores So most of the discussion up to now has been focusing on portability across ethnic and racial groups But we've also been involved in some some work looking at portability across different environmental strata within uk biobank and we find that actually Portability even within a racial group is You know it's not as high as one might expect So we think that genetic differences between groups are only one part of the differences that we see in portability And so that that's also two things first of all that you know, I think it's You know, even though even though we expect that portability is not going to be perfect between racial groups I think it's incredibly important to include different groups but secondly that I think that You know to the extent that we can capture as much environmental information on individuals as we can And also include that in the analysis of of you know, how predictive scores are will be very important Can I ask you to clarify? Can you explain a little more about the other reasons? That it's that the portability is low Yes, so um, yes, so so we have a project with molestivowski's lab at columbia and And in that project We we've been setting up different different strata of individuals in uk biobank for example by scs or or age groups or or sex like many like there's various different ways you can you can subdivide individuals within the the white british group and we find that there they Like polygenic scores that you develop in one stratum don't lift as well to other stratas as as they do In a cross-validation set within the same stratum. And so, you know, that's probably going to be a general principle of polygenic scores And you know, so so genetic differences between populations are also important I'm sure but that's not the whole story. And so it's important to think about other other kinds of environmental factors that may be Important. We don't have a good handle on those yet Excellent. Thanks. Yes, rafael. Oh my So this is in a way a follow-up question Can you say a bit about why what's what are the benefits of having a network as opposed to having wider distribution of of investigators doing similar A studies so if you instead of having what is it 12 or eight centers all and they all seem to there's not much geographical Viability Having many many smaller groups doing it and not necessarily coordinating at the at the throughout but more at the end or Organically, sure. I mean, I think we would want both ideally. So we really would like Lots of investigators to be addressing this question. It seems to be, you know, a major one On the other hand, we do need to have Some kind of a Systematic approach to this because it is something that you know in order to identify the the very top couple percent of people You need a large population to be able to do that And you also want people to be to be doing them in a sort of a similar or systematic way So I think there are two different ways to to approach this problem We're proposing the one that the institute would would initiate that would be a large network that would work together Probably have enough critical mass that it would actually bring the field along Which is what we've seen with Emerge and ignite and other kinds of programs But certainly want to investigate or initiated efforts to do this as well Follow-up is again relates to Jonathan's comment is that in general in machine learning problems Systematic approaches like this actually end up not working as well because the they all share the same bias by having it distributed And say different environments then that bias averages out, but if everybody has the same bias Then you you end up with a bias estimate sure and that's in no Understandable. I think in the from the machine learning standpoint, you're obviously no far more about that than than we do This isn't only the machine learning algorithm development aspect Which actually for the most part is going to happen outside of Emerge This is really the implementation You know, how do you how do you give this to a patient? How do you provide that information through the emr? How do you get physicians to follow it, etc? Which are our thornier problems So, uh, I think a mark was next and then Jeff What you do it So the The process of developing a new algorithm for eliciting a phenotype from the EHR is still very labor intensive And really a low throughput problem right low throughput process So for the the 20 diseases you're going to look at do you already have the phenotyping algorithms in place? Or will many of those have to be developed? And if it's the latter, I guess I'm wondering if you thought about trying to incentivize Innovation and making it faster to develop phenotyping algorithms Yeah, I think for many of the common diseases that are currently have you know have Polygenerous scores, we do have a phenotypes developed We wouldn't have them all and that might go into the selection of the diseases that we that we pursue It takes about a year in in Emerge to as you probably know To develop one of these so if there were one that had a really compelling Risk score, we might try to get the investigators to develop one of those But yeah incentivizing people to to develop other a phenotypes is something we would love to see happen I agree with you I was suggesting incentivizing Developing new methods to more rapidly develop the algorithms. Okay I think that's something we should yeah, we should look into I don't think we can do it And that's a different program from what we're proposing here, but but I'm glad that you're enthusiastic about it because we are too Jeff Otkin your next Couple of quick questions, so I'm going to go ahead and just tick them off and Hopefully they'll be quick What's the age-range you're targeting here? Are you including kids and if you end the elderly and Have you thought about the possibility of how risk scores may play out in the family context with Multiple members who may share a risk score and how that may influence behavior Secondly, are you only looking at conditions for which there are risk reduction? measures so-called clinical utility And then if you could just comment briefly on what the nature of the LC component of this is going to be Sure, so on your first question in terms of ages, we wouldn't have any restrictions on ages So so the full age range that the investigators are comfortable with recruiting and implementing on In terms of family members, that's a toughie and I I'm not sure that We probably wouldn't want to exclude family members I'm not sure that we would want to target family members me because of the reasons that you say that you know there would be some confluence of of Risk reduction recommendations, perhaps That would go from the parents to the kids or to the siblings or or whatever So so probably not that again Some of these are specifics that that the investigators really need to define and and propose and then and then be judged on that In terms of risk, sorry risk reduction. Oh the risk reduction recommendations I Would assume we would probably be on a firmest ground if we were using clinical professional guidelines and that's also gives us Much more likelihood of getting reimbursed by You know by insurance and that sort of thing. So I think that would be where the emphasis would be Maybe there aren't 15 of these but I think in two years There probably will be and you know, so maybe it's 20 or maybe it's 10 And then your second your third question on LC. I think you know You guys would be in a better position to propose the LC research questions that that could be addressed But I would think You know, what's the impact of telling somebody they're at high risk? What's the impact of telling them they're at low risk? How does that affect their quality of life? How does it affect their behaviors a number of things but others that you'd like to propose? We'd love to well in front of a practical perspective Will each of the centers be expected to have an LC component as part of their Um application or are you going to centralize that with one of the centers or how might that play out? I think we would let it play out as it as it plays out rather than you know I think that's one of the the areas of expertise that anybody who works in this area Recognizes they need to include some expertise, but having a separate LC component is probably not something that we would would require of each center Typically what happens in these programs is that we have an LC working group and they propose Studies that very often then we can come back to our LC colleagues and get additional support for which we love Or or do within the resources we have I've got wendy and then gale anyone else Wendy So terry in addition to the polygenic risk obviously a lot of the genetic risk is rare variants that are highly penetrant or moderately penetrant And i'm i guess it was probably because of cost that you weren't including that within the analysis But i'm just wondering because especially if we think about prs for Minorities and if that isn't coming up fast enough obviously many of these variants though for the highly penetrant genes still could be useful and if you were to try and Diversify in terms of using perhaps genotyping and sequencing based approaches You could cover sort of hedge your bets and perhaps get a more accurate total estimate of what risk is by including more things So i'm just wondering because i saw the cost of 350 dollars for the genotyping Which is generous in terms of what this would be and i would bet you could actually get a laboratory to do an Exome plus the genotyping for some with a clinical interpretation for something around 600 ish in terms of that and i'm just wondering if that were the case would that shift in terms of the approach I think you know being the genome institute We would love to capture as much genomic variation as we could Unfortunately exomes probably won't do it for us because polygenic risk scores are primarily based on giwas And the giwas 90 of the hits are not in the exome right But if you did an exome plus a genotyping array you could actually do both of those at a reasonable cost Would that perhaps be something that would be more comprehensive more forward-looking and given the huge investment You're going to make and you know getting this cohort together that kind of fixed cost You would get a much higher yield and it would be many of the same conditions You'd probably develop prs for and therefore have more positives that you could see what the impact was Yeah, I think that's something we could consider I think you know looking back at our strategic planning workshop that was held last month There was also a lot of enthusiasm for Genomic information outside of the DNA sequence and we want to you know Include enough room for that as well. So so I think this is something that's going to continue to change We will probably ask what we typically do in these is give Certain fixed costs to the investigators just so that everybody proposes on the same You know on you know the same playing field as it were But ask them also to say you know if you don't think this is the way to go Tell us why and tell us what else you'd like to come forward with Gail So Yeah, this is a lot to take in when not having been part of the planning group and so on and and I I really focused on your use of the word exploratory. So you said this is Really an exploratory project and yet I think Pat's comment and a couple from others is more but wait This is you know, you you'd like to be able to generate evidence that can only come through Some kind of rct or having a control group of some sort that's defensible So how could you explain to me how you you yourself are seeing this? As and I also excuse me, but I'm also thinking Back to the day the early days of Caesar when it was exploratory and then it was evidence based the second round And this is a lot of money and people and time and so on this is quite I was going to say I think this is more expensive for sure than the first Caesar. No, it's not actually Yeah, it's really not. Yeah, the first Caesar was about 65 million Oh, okay. Well, never mind that comment. Yeah, so it's a lot of money But the but just but the my my question about exploratory versus evidence based And how you you're thinking about that kind of investment and yeah, yeah, no, it's a good point I'm not sure I use the term exploratory. I think we were exploring if I did Yeah, I think descriptive is more more where we're going with this We would be exploring the you know the ability of the polygenic risk scores to predict the phenotypes and that sort of thing But but regardless of what what terminology you use We really wanted to propose a clinical trial and we just felt that we weren't a clinical trial To yeah, we just felt we're not there yet because of all the things We don't know yet and we're hoping that a study like this could you know Give us the information that we need to do a clinical trial in the future But but really at this at this phase with there's just too many gaps to be able to We do think it's going to give us some important evidence Though in terms of what people actually do when you tell them they're at very high genetic risk for a for a condition And that we think will be quite useful I never thought I'd be the person to say this but I think Introducing a pediatric population into this is going to make your hard job even harder Because it really depends. Are you talking about kids from age one to five? Are you talking about kids from age 10 to 15? You're not going to be able to compare them Across the adult cohorts very easily at all And so I think if you're starting to try to do the numbers of the number of traits and things like that I would just make sure you're modeling then the diversity of populations I think that's why someone else asked what age range If you want to get enough data to go across It's going to be difficult if some sites come in as pediatric. I mean, I know you've done that successfully in a merge But I don't think it was using exactly these kinds of measures And there are many fewer interventions and children obviously around things like diabetes and obesity there are But those may be quite different in a in a depending on the age of the child So you might want to at least think about particular age groups of children if you're going to include a pediatric and an adult population No, that's that's excellent advice. We we do recognize there are scores for ADHD Childhood asthma autism. So a number of childhood phenotypes It really depends to some degree on what the investigators are comfortable proposing And so we would rely on them to but don't they have to then agree across all the sites to conduct it I mean for the scores I mean, would they have to well, I'm just asking are you going to do an ADHD prediction across the entire So what would be the what would be the disadvantage of of doing that for for the children that they that they include I mean say you have some sites that you typically do that focus on adults and other sites that focus on children You're going to have some some confounding by you know One groups in Ohio and the other groups in Idaho and that that sort of thing But what would be the disadvantage? I just mean your ends are going to go down then So if you're looking at breast cancer, none of the pediatric sites are going to be able to do anything And if you're looking at Asthma and children then the adult sites might not be able to do anything. That's all I meant. Okay. So yeah, yeah No, that's that's very reasonable Jonathan I just want to reinforce something that the other Jonathan said said earlier There's a huge confounding between a lot of the environmental variables and ancestry as as we know So when you start to look at that kind of thing I think you have to be very careful that about that confounding and it may have a Significant effect on some of the sample sizes and some of the power that you have to to pull some of these things apart So as you go forward, I think that's something you probably should be very careful about thinking about Thanks. I completely agree Other questions or comments All right, I think that was a really good discussion And uh, well, I would like to move forward with a vote. Can I get a motion to approve the concept? Thank you, Jonathan a second Thank you all in favor Any opposed Any abstentions? Thank you, rafa. Okay. Thank you, terry