 Welcome, everyone, to the sixth bold predictions for human genomics seminar of this NHGRI seminar series. I'm Terry Minolio. I lead the Division of Genomic Medicine here at NHGRI, and I'm delighted to have the opportunity to share this seminar with you. Great. So this is, as I mentioned, it's the sixth in the of the sessions of the bold predictions for human genomics by 2030 seminar series. I'd like to thank everybody who's currently watching live and for the future people when this becomes archived, I thank you for watching it on YouTube. If you haven't already, you might want to watch some of the previous sessions that we've held the five previous sessions, especially perhaps the first session where Eric Green gives a wonderful introduction about details about the strategic vision and the context of the seminar series. This is one of many activities based on the 2020 NHGRI strategic vision that was released in October 2020 in nature, as you can see here. This series is highlighting the final section of the vision which features a sort of a new component that we added for this particular strategic vision. On 10 bold predictions for 2030, so we thought we would identify 10 things that we thought would be really cool to have happen in the next 10 years in genomics that we also feel are within reach, but perhaps a bit of a stretch. So all of the sessions in the series has now been scheduled. Each of the seminars is held on Zoom. They're all open to the public and their archive for future viewing. So the format of this session is like the previous ones. There are two speakers. Each of them gives a 25 minute talk and then there's a moderated discussion and there are questions and answers for questions from the audience and answers from the panelists. And we encourage people to enter your questions at any time in the Q&A box. In today's session, we'll be looking at the sixth bold prediction, which is the regular use of genomic information will have transitioned from boutique as it were to mainstream in all clinical settings. And this will make genomic testing as routine as complete blood counts. Our goal really is to sort of unpack this prediction over the next 90 minutes with two speakers. We have Jennifer Posey, who's assistant professor of molecular and human genetics at Baylor College of Medicine in Houston. And that then followed by Katrina Armstrong, who is the Jackson professor of medicine at Harvard Medical School and a physician in chief at the Massachusetts General Hospital in Boston. So Jennifer, take it away. All right, so I'd like to start by thanking the NHGRI, Eric Green, Terry Monoglio and Chris Gunter for the opportunity to speak to you all today about what I think may end up being my favorite bold prediction. And so I'll jump right into it. So bold prediction number six, this is that the regular use of genomic information will have transitioned from boutique to mainstream in all clinical settings, making genomic testing as routine as complete blood counts. And I'm thinking about this from a little bit of a kind of a stepping back and thinking about it from a bit of a high level. I wanted to first start with where do I think we are now just in very broad strokes. So 2021. A lot of the genetic information that's used is used in very limited clinics, either genetics clinics or clinics where there are clinicians who have some genetic expertise or or some comfort level with applying that type of testing and utilizing the results. And the reason is perhaps most often really that there's a specific diagnostic question that's being asked for that testing and the how it's often a one in done sort of a one time test or one time data analysis. The results report is generated and the patient and the clinician move on. But thinking about where we could be in 2030 is this bold prediction is correct. I think that, you know, the genomic information will be readily available in all clinics and all clinicians across both general and specialty expertise will be using this information. It'll be used not only to answer particular diagnostic questions, but also in a preventative way to address future health risks. And also, I think that we may see iterative data reanalysis and iterative testing so that it's no longer simply a one in done type of test for many individuals. So what are some of the achievements that I think are needed to take us to 2030 and to take us to achieving accomplishing this bold prediction. Certainly public private partnerships. So looking at addressing infrastructure needs, healthcare adoption and diffusion. Continue genomic discovery to inform clinical care and piloting mainstream implementation. And of these for the top two, Dr. Armstrong is going to be addressing in the next talk and I'm going to focus my talk first on continue genomic discovery and then move on to piloting mainstream implementation. So just to give you a little bit more background about what I mean by these two items. One of the questions that I'm very focused on in my research lab is how can we even more comprehensively understand how pathogenic variation rare variants that two or more genetic loci impact a person's health. We're very good in many ways at understanding when a person has one gene that's not functioning in the way that we anticipate what the health income of that might be, at least for a large subset of genes. But when you start to involve two or three or more genetic loci or genes that are no longer functioning in the way that we anticipate the outcome of that becomes much more complex to predict. And yet we need to be able to do this for our patients and we need to understand the biological implications of these types of scenarios. With that in mind, one of the questions that our group has begun to work on is whether or not we can computationally model gene and variant relationships and phenotype relationships to begin to understand this even better. And then on the right side, I'm going to be addressing a little bit how we can deliver genomic information to specialty clinics in a way that's clinically informative for clinicians and patients. And how we can best partner with our non-genetics colleagues to facilitate their use of genomic information more broadly. And so we'll jump into the item on the left, the topic on the left to start. In many ways, I think we often think about the relationship between a person's genetic information, the sum total of their genetic changes, and the impact on disease expression or their clinical presentation, their clinical features, is really a one-way street so that if you have a genetic change, such as we'll say the PMP-22 duplication, which is the most common genetic form of Charcot-Marie-Tooth neuropathy, that you would anticipate that person will develop if they've not yet already a peripheral neuropathy. So really that's using the genetic information to predict a very discrete health outcome. And yet I want to give you an example of a young girl that I saw here many years ago now who taught me that that relationship is not quite as simple as perhaps I used to think. So I first met her when she was about four months old. She was hospitalized for failure to thrive, developmental delay, and low muscle tone. She had two older brothers who were in good health. And by the time our team was consulted, the genetics team, she'd already had a chromosomal microarray. You can see the results of that down at the bottom of the screen. And in red, what I'm trying to show is that she had a loss of X chromosome-specific probe. So she had monotomy X or Turner syndrome. And at the time, we felt reasonably confident that this explained some of her failure to thrive, her developmental delay, and her low muscle tone. But things started to change when we saw her in follow-up in clinic. By the time she came back to clinic several months later, she had much more severe developmental delay. She had very severe microcephaly or a small head size, and her growth had slowed even more. In fact, it was much more severe than what one might anticipate for Turner syndrome. And so we went back and we thought about the genetic information we already had. And we decided to get additional data. We decided to get a karyotype. That karyotype, or high-resolution chromosome study, showed on the right that she actually had two populations of cells. She was mosaic for monotomy X, as well as for triple 80. And so, in fact, her situation was much more complex than what we initially were able to determine at the time of her first consult. And what this taught me is that there really is a virtuous discovery cycle between a person's genetic information and their health. And this may evolve over time. And so that relationship between genetic information or finding a variant and then thinking about that clinically and understanding whether that explains the clinical picture, that should happen continuously over a patient's lifespan. It's not a one and done. And it has to happen in close partnership with the patient and, of course, the clinicians that are seeing the patient. So in thinking more about the relationship between genetic information and the clinical expression of disease, there are several places where more work is needed to be done before 2030. The first, of course, the biological functions and disease relationships of all 20,000 protein-coding human genes need to be elucidated. And there are different rare disease research groups, including the Centers for Mendelian Genomics, which we have been a part of that have made great strides in identifying novel disease genes. And you can get a sense over the first eight years of that program what the discovery trajectory has been. So we have that program alone. And we know that future programs such as the Mendelian Genomics Research Center will continue in that discovery vein. But much more than just understanding the relationship between a single disease and a single gene. We know that we need to understand more about the influence of different variant types on gene and protein function and their ultimate impact on health. Not every variant in a given gene is necessarily going to have the same outcome on health. And then one of my personal interests, we need to better understand the impact of combinations of variants in more than one gene and how that can influence a person's health. And in many ways, I view that as a bit of a stepping stone to beginning to understand how the combination of a person's genome and its entirety influences their health. Because no single one of us is the sum total of one or two or even three of our variants. It's much, much more than that. So I want to talk to you a little bit along those lines about dual molecular diagnoses. And what I've shown you here on the left is an individual who has two conditions. R-SCOG-SCOT syndrome caused by pathogenic variants in FGD-1 and, listen carefully, caused by a pathogenic variant in a gene called P-A-F-H-1-B-1. A bit of a mouthful there. Two genes involve two different disease traits or two different molecular diagnoses. And this results in a blended phenotype. So when this person comes into clinic, they don't necessarily have the two diagnoses kind of tights across their head. You see an individual that may have, in this case, a combination of clinical features that you might not have seen before. And it can be very difficult to ascertain that this is actually a person who has two different things going on. Accurate diagnosis, however, is critical because it informs surveillance and management for this particular patient. It helps us tell this patient's family what the recurrence risks are. Our other family member is going to be at risk for either or both of these conditions. And, of course, it reminds us, just as with the first patient that I told you about, that our diagnostic odyssey may not end with the first molecular diagnosis. We always have to go back to the patient's phenotype, their clinical presentation, their health status, and revisit their genetic information in that context, which is going to evolve over the course of their life. So we've become very interested in understanding the frequency of multiple molecular diagnoses. And in partnership with the Baylor Genetics Diagnostic Laboratory, we set out to look at an analysis of sequential diagnostic laboratory referrals for diagnostic exome sequencing. What we found, and this is for non-cancer indication, what we found in this particular cohort, their molecular diagnostic rate for exome sequencing was about 28%. But when we looked at the individual who received a positive test report or a molecular diagnosis, 5% of them actually had two or three, one even had four molecular diagnoses. So while dual and multiple molecular diagnoses were known before this, what we learned from this study is that they were actually more frequent than we ever thought they had been. And in fact, this almost certainly represents an under-acertainment. So fast-forwarding a few years, I want to show you some data that was led by Pengfei Liu again in collaboration with our diagnostic laboratory. He looked at two previously reported cohorts of diagnostic exome sequencing data and performed systematic reanalysis. And what he found in both cohorts on the left and on the right is that that reanalysis identified additional molecular diagnoses that had not been identified at the time of the initial reporting. And those are the red portions of the upright bars that you can see here. Novel gene discovery drove a majority of those new molecular diagnoses. What excited me the most about his data, though, is that there was an almost doubling of the number of dual diagnosis cases by reanalysis. And these two cohorts alone, that number went from 25 to 48. What's the other challenging aspects of dual molecular diagnoses from a clinical standpoint is that they can come into sort of opposing categories. I mentioned earlier the example of a person with a distinct phenotype that might have a set of clinical features that you've never seen in combination in a single individual before. And you might think that they had a new syndrome or a new condition. The opposite can also be true that a person might have a set of overlapping phenotypes where they might look to simply be more severe than what you might expect with one condition or the other. And again, clinically, this is very difficult to ascertain unless you're performing unbiased or very large genetic testing such as an exome sequence, whole genome sequencing. One of the questions that we wanted to ask was whether or not we could begin to computationally model these two opposing categories of blended phenotypes, the distinct and the overlapping. To do this, we took 80 cases, 80 individuals from the diagnostic laboratory who had dual molecular diagnoses, just exactly two. And two physician scientists classified them as either distinct or overlapping pairs of diagnoses. We then went back and we utilized human phenotype ontology terms, HPO terms, to generate a phenotypic similarity score, which was essentially a comparison of the phenotypes of those two molecular diagnoses. And indeed, when we did this, we observed that the individuals with distinct disease traits or distinct phenotypes had a lower phenotypic similarity score than those with overlapping. And this was very exciting in many ways. I think that this represented an early step for us in how we think about integrating phenotype ontology terms into some of our analysis. So from there, we started to ask many more questions. First, we wanted to know, could we take advantage of the ontological structure of the human phenotype ontology, the HPO, to perform pairwise comparisons of patient phenotypes. We wanted to see the ontological structure of the HPO, just a very small portion of it depicted here in the bottom right. But what we were really doing is taking two individuals and their list of clinical features and comparing them and generating a phenotypic similarity score. We also wanted to ask, what can we learn if we apply this approach to a single genetically heterogeneous condition, so one condition. We started this by looking at Robonaut syndrome, and this is a skeletal dysplasia syndrome that has a very characteristic clinical findings. And so many individuals with Robonaut syndrome will receive their Robonaut syndrome diagnosis, their clinical diagnosis, in advance of genetic testing, that high clinical suspicion. And yet when you perform the genetic testing on them, you see that there are multiple different genes that can be responsible for their Robonaut syndrome. In this case, we looked at 68 individuals with Robonaut syndrome, and in this particular cohort you can see the breakdown of the different genes that were responsible for their Robonaut syndrome. So again, genetically heterogeneous. And with that pairwise, patient to patient comparison, generating phenotypic similarity scores, and showing these data on a heat map for the individuals with the most similar phenotypic, with the highest phenotypic similarity scores, the pairs that were most similar were clustered together. So this is a cluster analysis. And what was amazing to me as a clinician is that despite the fact that all of these individuals had Robonaut syndrome. This is a cluster analysis by phenotype alone, no genetic data, genomic data was included in this cluster analysis, yielded these gene specific clusters. And so you can see here in blue, a very large cluster that contain largely dvl one and dvl three variants individuals. In the bottom right, you can see another smaller cluster that are individuals with truncating variants in a gene called frizzle two. In the middle here in purple, too much smaller clusters of individuals who also had frizzle two variants, but they were different variant type again going back to what I said earlier about how important it is to understand variant type impact on phenotype. These were missense cluster. And so this is absolutely amazing and it tells us that we can really begin to computationally understand human phenotypes in a way that perhaps is even even more. So then we're able to do in the clinic when we're seeing one patient at a time. Right next, I want to show you an example of how we used the phenotype similarity score analysis in a slightly different way. In this case, we studied this is a single case study, a young boy with a very severe syndromic developmental delay. And the comparison that we performed here was the patient, this young boy, to several different disease traits. What drove this is that when we performed an analysis of this exome sequencing data in the research setting, we identified that he had rare pathogenic variants in four genes, VC4H2, NAV2, MUSC, and CAPN3. Three of these were already associated with well-described disease entities, but the fourth, NAV2, actually had not been described in association with human phenotypes before. And so the question was really, can we begin to dissect the contribution of each of these rare variants to this individual phenotype? And what can we begin to try to learn about how NAV2 might be contributing? It's especially difficult when you have multi-locust pathogenic variation and you have a novel disease gene. We did a very similar analysis to what I just showed you before. And in this case, the graph to the way that I'm displaying the data is a little bit different. We took the patient's phenotype, which you can see listed across here, and shown in the red box here across the top. The patient had all of these phenotypic features, and we looked at which of those features would be explained by each of the genes, by pathogenic variation in each of the genes that were identified. What we found was that this particular patient's phenotypes were incompletely explained by rare pathogenic variation in the three established disease genes. Yes, they explained parts of the phenotype, but not completely. And yet, for NAV2, we were able to go back and integrate mouse phenotype data, and this enabled a computational assessment of the phenotypic contribution for this candidate disease gene as well, even though it's never been described in association with human phenotypes before. And so we really were able to begin to use this type of analysis to convince ourselves that this individual truly had multi-locus pathogenic variation with four genes, each contributing different aspects of his phenotype. Now I want to switch gears just a little bit and talk more about genomic information in the clinic and how we can deliver this in a way that's informative for clinicians and patients, how to best partner with non-genetic specialists. The first study that I want to show you is been led largely by the Human Genome Sequencing Center and David Murdock, who's also a colleague of mine in the clinic. It's called heart care, and this involves piloting the implementation of genomic risk assessment for cardiovascular disease. So this is implemented in cardiology clinics, and importantly, part of the multidisciplinary team that developed this particular study and this particular test involves not only geneticists and clinical geneticists, genetic counselors, but also cardiologists, the individuals who are going to be delivering the results of this testing themselves. It's important to know that heart care is actually at no cost to the patients, no cost to their insurance companies. This was fully supported by a Baylor, and there was early engagement of key stakeholders, the patients and the cardiologists in particular, and the results were integrated directly into the electronic medical records. The heart care report contained four different components. The first were actionable cardiovascular disease findings, so pathogenic or likely pathogenic variants and 158 different genes. These were genes that are associated with arrhythmias, cardiomyopathies, aortopathies, dyslipidemias and others. Patients would also receive a polygenic risk score if they were in the top 5% of risk for cardiovascular disease. They would also be told if they had high risk alleles in a gene called LPA, lipoprotein A, and then finally, if they had any pharmacogenomic findings that might influence their metabolism of certain drugs. I want to show you just an overview of where we are with that study, so 709 patients have completed heart care. And in 64 of them on the left, they received actionable cardiovascular disease diagnoses, and I've broken this down for you by clinical category, 50% involved dyslipidemia diagnosis, 37% of cardiomyopathy diagnosis, and some also received an arrhythmia or aortopathy diagnosis. But of course, that was not the only form of result that was available. 9% of individuals looking on the right received information about a high polygenic risk score, 20% received information about a high risk LPA variant, and almost half received information about a pharmacogenomic variant. And so not including the pharmacogenomic variant, the overall cohort positive rate was 32%, 32% of individuals that received one or more findings, not including pharmacogenomics. And so this was really, we thought, a very high yield study. We wanted to then build on this, and what we're currently planning, which will soon be underway, is a cardiometabolic precision health pilot in South Texas. This is going to be built upon the heart care pilot, and it's a scalable population based precision health pilot. It's a collaboration with doctors hospital at Renaissance, the DHR health system. And the first step has been engagement with two different clinics, the cardiology clinic, similar to what we did with heart care, also diabetes clinic. So the endocrinologists who are treating individuals with diabetes. So the results report that patients will receive will include information about both cardiovascular and diabetes disease risk. One important aspect of this pilot as well is going to be integration with Project ECHO. And this is going to allow us to engage in virtual peer to peer learning with our DHR colleagues in South Texas. Case co-management between Baylor and DHR experts, and effective delivery of boots on the ground clinical expertise. And I want to give you one final snippet, one final example of how I think genomic information can be very helpful in a non-genetics clinic. This particular example actually comes really from a transplant clinic. So I want to tell you very briefly about a donor who is almost 17 years old, an organ donor. He was diagnosed with and passed away from a brain tumor called pleomorphic xanthrolastrocytoma. There were four recipients of his organs, which you can see who you can see listed here shown in blue. And unfortunately, all four recipients ultimately went on to develop cancer. Actually, not very long really after their transplant. The origin of their tumors could not be determined by standard methods, including pathology. However, we were able to use, the Genome Center was able to use somatic mutational profiles in the donor, as well as three of the four recipients. And this told us a couple of things. First, that the donor's tumor was an aggressive tumor, unexpectedly for its tumor type, aggressive. It had a high potential for invasion and metastasis. And this is based on what was seen in the genomic profiling. And second, it became very clear that the organ recipient tumors were descendant tumors from the donor's tumor. What this tells us is that in the future, genomic profiling of a donor tumor prior to transplantation may be used to predict its potential for invasion. I'm sorry, genomic profiling of a donor organ may be used to predict the potential for invasion and metastasis if that donor has a known tumor. In this particular case, although the donor's tumor was known to be sequestered in the brain, imaging did not show any tumor elsewhere in any of the transplanted organs. And yet this still happened. The role of genomic profiling to really use as part of the assessment for transplant safety is something that should be considered. And with that, I want to go ahead and wrap up. I think there's really a need to continue to develop methods to understand and predict the health impacts of multi locus pathogenic variation. I hope I've convinced you that the relationship with patients and the clinicians that are on the receiving end of the genomic information, that relationship is going to remain critical as we move forward with this. I was able to show you an example of a precision medicine health pilot in which clinically relevant findings were identified in 32% of individuals for cardiovascular disease risk alone. And we're now expanding this to South Texas populations and integrating peer to peer education. And then finally, I hope I've shown you a small example of how molecular pathology profiling can potentially be used to reduce the risk of transplant cancer transmission. And so now that takes me back to our bolder predictions and I want to build a little bit on bold prediction number six. There are three things that really came to mind when I thought about this prediction. The first and I mentioned this a little bit at the beginning of my talk is that I think the focus of most clinical genetic testing is going to transition from diagnostic to largely predictive and preventative. I expect that in 2030 we're going to have many, many more clinical guidelines in place that will have will detail recommendations that are tailored preventative therapies for identified disease risk. I also think that reanalysis of existing genomic information, harmonization of multiple data types and sources for people who may have had genetic testing in different places or different types of tests. This will become routine. It will become routine clinically. And finally, I think that the information used broadly in clinics across all clinics is going to include more than genomic information. Other omics modalities such as transcriptomics or metabolomics will be used and they're going to provide important snapshots of particular tissues or time points during a person's life. And I want to certainly acknowledge there's several many team members here who contributed to some of the data that I was able to show you today. And of course it goes without saying that I have many, many more tremendous colleagues and collaborators whose work I didn't mention today. And of course, I'm also very thankful to the NHGRI for my K-O-8 award, which really got my own lab and my own career off the ground and running. And with that, I'll let Dr. Armstrong continue. Dr. Posey, thank you so much for that fantastic talk and I want to thank the NHGRI leadership and everybody here today for allowing me to speak about this incredible prediction, which I will say for me is maybe a little bit more of a pipe dream that I've had for a very long time. I'm going to go ahead and show my slides. So I'm here to talk about prediction number six and build upon Dr. Posey's amazing discussion about a key issue that I'm going to come back to. The bold prediction that the regular use of genomic information will have transitioned from boutique to mainstream in all clinical settings, and I'm going to come back to that, making genomic testing as routine, as complete blood counts, and I'm going to come back to that also. So the first thing I want to say is that when they asked me to talk about this, the first thing I thought of is that it's a terrible, terrible idea to talk about predictions. Because the reality is that nobody ever remembers if you're right, but it is famous when you're wrong. Some of the most famous errors that have been made and prediction here, as you see Thomas Watson, chairman of IBM saying, I think there is a world market for maybe five computers. And so when I think about predictions, it's not that I want to go down as the person who said there will be at the most five computers or who wants to hear the actors talk. So I think my favorite quotation about predictions is that if you're going to make a prediction that the best way to predict a future is to create it. So if NHGRI has gone out and made these predictions, they've now got to create this future and I think there's nothing more exciting than being part of that. So if we're thinking about that prediction, what is going to make that happen. I think there are two key steps that will make that bold prediction happen. So the first is that there are accurate, affordable and available genetic tests. I think we've seen that happen incredible progress and I'll talk a little bit about that largely driven by NHGRI. Look at the work that Dr. Posey and many, many others are doing bringing those tests to clinical medicine. The key second step here is that those tests have to diffuse into clinical practice. We have to be using them to make that prediction happen. So I want to talk for a second about a concept that I've studied for a long time in my lab and it's used in many places so called the diffusion of innovations. What is it that means, I bet if I were to ask everybody on this phone, on this Zoom, I always forget where we are these days, call Zoom, whether or not they have an iPhone that pretty much everybody could pick this up, that we all sit here with an iPhone. So what makes some things diffuse, some technological advances diffuse across and what makes some not diffuse? Well, so Everett Rogers sometime ago should essentially show that essentially all technology diffuses and then S-shaped curves starts out slow. There's a slope that picks up and then it flattens at some space. And that curve seems to be determined by the fact that we can group people or group users into these five main categories. Innovators, the very early group who use some form of technology, early adopters, early majority that comes next and then the late majority and laggers. You guys can probably think even now about people that you know and where you might group them for different technologies in your own either home life, work life or scientific life. So if you look at this concept and we asked ourselves if we're thinking about genomics, how long, so 2030 we've got, we're sitting in 2021, how long does diffusion normally take? What you can see here is that there's actually a dramatic variation in the time that it takes a technological innovation to diffuse across a population. So here I think some of my favorite ones are the clothes washer. I think you can see here that took actually for a very long time 1930 up into 2000. Whereas you can see if you look to the right of your screen, internet, cell phones, rapid diffusion. So going from almost nobody using the technology to essentially everybody using the technology in 10 years or less. Similar stories can be told for healthcare technology. So if we look at a healthcare technology, it generally follows the same S shaped curve. But the reality is that in healthcare and this is true outside of healthcare also, most innovations are actually never widely adopted, and we have to make effort to actually get them to the top of that S shaped curve. It's also true that many are eventually discontinued something for us to think about as we move forward in a genomic medicine era. Just some examples from medicine itself. So this is for a neonatal intensive care unit. And it turns out if you look at Nick use and what happened in our country and Nick use that it took 20 years for hospitals go from no hospitals the first Nick you in 1960 to 5% of the hospitals having Nick use across country 20 years. And then in 14 years they hit that curve and went from 5 to 30%. If you look here at MRI diffusion and turns out that MRI went faster went from no hospitals to 45% of hospitals in 14 years. If we think about something that honestly many of us are very common very associated with the use of statins. Here's data that actually is only from 2002. But if you look back before that, if you look at eligible individuals that went from zero to 50% of eligible individuals, beginning a statin in 10 years, and then it took another 10 years to go up that 10% from 50 to 60%. So what do we know about this prediction now for genomic medicine? Well, the reality is not all that much. But I'm going to focus for a minute on something I've studied a lot over my career, which is the use of BRC one and two testing and women who are eligible or considered eligible for testing. So here's the genetic testing or the BRC one and two diffusion timeline right now some key events I kind of marked here for you. Discovery of BRC one and two in 1994, the first commercial test in 1996, several major groups were back amending testing in 2003 2013 and then with the change in the patent law multiple companies now offering testing over the last several years. So what do we know about the uptake what do we know about whether NHGRI is on track to make this prediction happen. So it turns out here I'm showing you data both for BRC one and two testing but also for genetic testing for developmental delay in honor of Dr. Posey another area that we have considerable consensus about the use of that genetic testing. The reality is that genetic testing has not yet hit that fast uptake part of the curve that we expect to see when it takes off and will get us to the 2030 prediction of it being commonplace across all of our clinical areas. So most of the data that come from true population based surveys would argue that only about 20% of eligible individuals are probably getting genetic tests if either for hereditary breast and ovarian cancer or for developmental delay. I did point out that there's some several studies here I can see in the dots that have higher estimates and that if you look at the curve it's definitely up. I'm an optimistic at glasses half full person and I would say this curve is going up and that we are headed in the right direction. But the reality is that we need to make sure that genomic testing does not become stuck at that early adoption 20% of users 20% early adopters that we saw in the study a long time ago looking at early adoption of the RCA one and two testing and it turns out that the people who use it early are different and the factors that drive their decisions are different. So when I'm an early adopter and this has been shown across a number of different areas not just genomics. When you have an early adopter they're often very innovative and driven by that level of innovativeness they're driven by a sense of compatibility with that testing and barriers are less of an issue so barriers such as testing complexity that perceive benefit of the testing have less impact on that early adopter group. So for us to get to that 2030 goal the research suggests that we're going to have to change and do more to overcome those barriers to get past the early adoption. There are three domains that investigation efforts can actually or can be used to move this forward. What is it that will move diffusion beyond early adoption. I want to focus on these three areas because they're often discussed as ways that we will achieve this grand vision. How will that happen. The private sector. The health care sector and the public sector. So often in most models of diffusion the idea is that we create a new test or let's say we create a new drug and then we rely on the private sector to achieve the diffusion of that into our clinical practices or to our patients. And certainly there is a lot of that going on in genomic testing right now. Enormous numbers of market entries into the country's companies starting new genetic tests on the market. This is an article from Catherine Phillips that came out in 2018 that looks at the entry of new genetic tests onto the market. And this is not the ancestry. These are actually as you see on the right the clinical data that are being developed and put into the market here for prenatal testing hereditary cancer testing oncology diagnostics and treatment. So an enormous effort by the private sector. This just shows one of these many graphs that is around that shows the large number of patients or individuals up into the 25 30 million who use some of these online genetic services here primarily around ancestry but significant proportions of individuals. So I think one of the questions is how much will this drive that old prediction is this going to solve it on its own. So I will say I am skeptical. I do not think that we're going to achieve prediction number six simply by relying on the private sector alone and I don't think that for several reasons. The first is that those efforts have remained largely unconnected to health care. They depend upon patients to seek out the test and to use test results effectively. Lots of efforts to try to fix that. But that model right now places a relatively large burden on patients. And that burden those issues can create multiple challenges in implementation. One of the earliest studies we did actually looked at what happens when you actually offer patients their risk information both to genetics breast cancer risk models breast cancer genetics mutation risk. This just as a graph of what we a study where women got given their genetic information got given their risk information. That's what's on the left axis of these graphs. And then they were asked two weeks later to repeat back to tell us what they thought their risk was. And I think you can see there's just a wide wide range from what the information that was conveyed to what was actually taken home. So even risk communication of basic information outside of the health care environment outside of a physician relationship is really challenging. And I think we'll make this paradigm in its ability to achieve that bold prediction. The reality is also that we're discovering that there's decreasing use. This is just some data online data about the decline in DNA kit sales. So actually a drop in the rates of people ordering these online testing. That doesn't mean these companies don't have a lot to offer. That doesn't mean that there isn't a big part of innovation happening through these particular companies. Now come back to that. But my argument is that if NHGRI if we're going to achieve this bold prediction it is not going to be based on the private sector alone. So I want to share for a minute some work that we've done focusing on health care. And what is it that is driving the utilization of this type of test within health care. Again focusing on breast cancer and breast cancer genetic testing. So health care drives the adoption of most health related services in our country, including diagnostic and predictive testing. But what drives the use of genomics in health care. So we did a study some time ago of about 3000 women with breast cancer between 40 and 65 in Pennsylvania and we also surveyed their doctors. So we surveyed their medical oncologists and surgeons and linked those data. We asked them about the use of BRC1 and 2 testing and also about oncotype testing collected a lot of eligibility information and validated it through chart review. We found this again and again as have many other labs that about 40% of women in this sample again it was a selected sample of high risk individuals reported BRC1 and 2 testing 39% women here reported oncotype DX. It turns out we measured a whole slew of things that were patient driven attitudes that you would argue we're going to drive the uptake of this testing. We measured attitudes about whether the testing would be helpful in managing cancer risk, whether it was too expensive, where there might lead to insurance problems. And you can see the proportion of women in the study who endorsed one of these attitudes. But the point I want to make here and this is where I think the key is in our health care investigation that needs to happen is that these attitudes did not determine whether or not the woman had undergone testing. So what determined whether the woman had undergone testing was whether or not their health care environment their surgeon or their medical oncologist had recommended testing. And there was an enormous amount of that that drove the use of these new technologies. And that's part of the reason I will say that I think the private sector solution alone is not going to solve this we have to engage the health care system. And the provider is for us to be able to reach that bold prediction of getting this used in clinical settings. This just shows the differences and women who by whether they underwent BRCA testing by whether their surgeon recommended it. Again, these are women with breast cancer or their medical oncologist recommended it, and you can see those are huge effects, huge, huge effects that have been seen in multiple other studies. And then if we look at the physicians what made them recommend it so we now have data from those docs linked to their patients. And what we can say is that actually when we now look at doctors we have a different story that actually their attitudes about the testing were very predictive of whether or not they recommended it to a given eligible patient. If they thought that BRCA one or here I'm showing you the data for oncotype also was helpful for patient management they were much, much more likely to recommend it. If they thought it was too difficult to arrange they were less likely to recommend it or too expensive as you can imagine less likely to recommend it. So what have we learned so we've learned that a lot of achieving that and bold prediction that is going to come from engagement with the health care sector, and that the health care sector what drives use in health care sector, and this gets back to our diffusion of innovation model is really some big areas so relative advantage what is the relative advantage of using the genomic test over other strategies. I will just point out that an enormous amount of what Dr. Posey and what I'm going to come back to here is discussing is is the demonstrating the relative advantage demonstrating in an invisible way that can make a change the adoption curves. Infrastructure needs are critical absolutely central to whether or not that physicians health care providers are going to be able to adopt this not news, but absolutely central to reaching that goal compatibility with current practice is a critical issue. But I want to call out two issues that I think have received less attention so far and come up a lot when we talk to the physicians and then other studies we've done. So the first is that actually a critical issue in the adoption curve is the presence of opinion leaders and the norms roles and social networks within the practices. This has been shown for many, many different diffusion of technology, but an absolutely central issue as we looked within our own data from our study that I just showed you, as well as others have seen and many other studies of the diffusion of technology focusing on genomics. And this I think is a critical issue that has received insufficient attention if we're going to achieve that bold prediction. The last I just want to point out is that it's very hard to innovate it very hard to implement something that changes all the time. And it turns out that one of the challenges as we face forward and implementing genomic testing is us coming to a place where we understand how can we create a model of what genomic testing is. How can we study and actually determine what that effective model is so that we can actually move to the stability of an innovation that can allow us to implement it and reach that goal. And as I said, Dr. Posey has spent a lot of time talking or her talk really focused on this critical issue of relative advantage and how the continual implementation can just can actually demonstrate that relative advantage. I want to spend a couple of minutes on a pilot project that we've been doing this really been focusing on these three other issues at Mass General. And this is just an effort that we started over the last several years to try to understand what can we do to achieve that NHGRI bold prediction just within my world here. If I can get it done here for you Terry maybe that counts. So if we can just do it here how would we do that. We created an MGH genomic medicine service I know that Dr. Heidi Rem is going to be visiting and talking with you about a bold prediction soon so led by Dr. Rem with a huge slew of other people with this basically follows the opinion leader model. So we have a central core support service that provides the infrastructure that we talked about. But then we have opinion leaders leaders in each of these areas of genomics champion and they come together and monthly joint meetings to create that social network and the norms about the use of genomics within their group. There's been lots of outcomes of this that I won't have time to share but I'll just say one of them was recently published in a paper in a kidney journal just looking at the clinical studies successful kidney genetics clinic and even this has now led to downstream positive effects on the other clinics adopting and disseminating new technologies and genomics simply because of the opinion leader and social norms that we can create. So I want to point out that when we do that we can do it in an evidence based way and I think this is a critical area for research as we move forward. It turns out we can design our evidence based genomic medicine and we can use strategies that help us to understand how to do that. I'm going to talk to you just about one method that is used in other fields to predict consumer decisions which is called con joint analysis or discrete choice experiments. In that model what we do is that we identify the modifiable attributes of a genomic medicine service. In this particular study we looked at where would it be delivered. Was it in a primary care provider's office or in a specialist's office. We actually looked at race specific versus non racially specific marketing because of the hypothesis in this study. And we looked at issues about confidentiality and insurers having access different we use a fractional factorial design different individuals get different scenarios and then rate the likelihood of testing and it's been shown to have a much stronger correlation with subsequent behavior. Here what I can just show you and I'm just going to show you one piece of data from that study. But it turns out that if we change the attributes of the test if it's racially specific marketing if we have a lower likelihood that patients will undergo testing if we obviously if we show insurers the results a lower likelihood but also if it involves a specialist visit. We've now done other studies like this and actually working now on a series of studies actually out in the South Dakota with Lakota tribes about how to use this approach to better design health care that actually reflects the priorities of the tribes. So using evidence asking the questions about how we can achieve this bold prediction. I just want to finish by talking for just a second about the role of the public sector, because what has been remarkable over the last years of NHGRI has been to see the extraordinary commitment by the genomics community to ensure that genomic research and data is a public good. A public good as you all know is defined by the inability to prevent access to that good. We all can gain access and by one person's use not affecting another person's use access to clean water and air scientific knowledge. I will say many people are now calling for COVID-19 vaccinations to be a public good. And I just want to say this issue is essential if we're going to achieve that goal to continue the public sector view of creating a genomic public good. But I think there's more opportunity here for leadership from the public sector and for public private partnerships maybe like we've seen with COVID-19 because there's several infrastructure needs that require large scale collaboration across health care entities to achieve this bold prediction. We need to have a public-private partnership around integration into electronic medical records, processes for interrogation of existing genomic information maybe on our smartphones like we're going to hear an unlader bold prediction. And I'm just going to say absolutely critical for access to genomic services among disadvantaged groups. I've spent a lot of my time studying disparities in genomic medicine. I'll just show you one piece of data that continues to show that when we look at uptake of genetic tests that we have uneven uptake. Uneven uptake by socioeconomic status, uneven uptake by race and ethnicity. Here BRCA testing among women with early onset breast cancer significantly lower among black women than among white women, particularly if they're high risk. So a critical issue that requires public sector leadership to actually undertake if we're going to get to that bold prediction. So I'm going to end by saying that I believe that with these pieces with the private sector innovation with investment in research and in in terms of designing healthcare delivery understanding how to influence new models. And with the public sector taking a lead on the public-private partnerships that I do hope that we're going to get to this bold prediction of the regular use of genomic information. I'm going to pause for a second that as I was saying that I realized that there's some parts about this prediction that I would like to change. And so for if I were writing this prediction I picked up two things that I would have changed. So first is all clinical settings and the second is routine. And so I want to call out what I would say is maybe something that we as a genomics community can do better. So when we look at testing right now in healthcare there is widespread misuse of testing. I often say the last thing we want to do is rebuild a broken system. So tests are ordered when the data already exists, tests are not ordered when they would change management, test results are misinterpreted, false positive results lead to downstream costs and complications. And so when I read that bold prediction though first thing that I thought was please no more unnecessary complete blood counts. I don't need more of them. So I'm going to put an asterisk and say like if I have a pipe dream today it's not that we're using it in every clinic all the time. It's not like we're ordering a morning genomic test like we order every AM a CBC is that genomics will have created a new paradigm for the use of predictive and diagnostic testing that obtains and integrates appropriate new information with existing patient data at the point of care. And genomics has the ability to do this which no other test does because we can re interrogate the information to provide new data new help for a clinical decision. So I'm going to end by saying if I were looking forward now we talked about our adoption and diffusion curve and the research and the data that suggests how will we get to that bold prediction. But I'm talking about another curve just for a second as I finished which is called the Gartner's height curve. I think we've probably all seen this but the general idea that we start with some technology trigger we have a peak potentially of expectations. And then we go through sometimes what we might think of as a trough of disillusionment on our way to the slope of enlightenment. And so I would say if I see where we are in this bold prediction heading for the next 10 years I think we're headed up that slope of enlightenment. We're getting to the implementation of genomic testing in so many of my practices right now in our nephrology practice in our GI practice in our ID practice in our primary care practices. But I think if we're going to continue that slope of enlightenment there's some major efforts that will make it work. So first I think it's time to move to large scale clinically embedded research consortia that can continually generate evidence about the relative advantage of genomic tests not one stop one you know stand up things but where we can understand about these things and major clinical areas heart failure lung disease kidney disease because by demonstrating that relative advantage we're going to create the information that will create that adoption. As I said I think public private partnerships are critical to develop and evaluate new models for genomic medicine infrastructure needs, including counseling models incredible innovation going on and models of counseling and education right now. Amazing opportunity. And then finally I will just say as we heard Dr. Posey talk, one of the most important things that we can do to achieve that is to continue to invest in the pipeline of clinical investigators through KO a K 22 and K 23 programs, because they will be those opinion leaders in the clinical areas to create those social networks and it turns out that without that, it's very, very hard without those opinion leaders in the clinical areas to get to that bold prediction. I have a clue of people here from the centers that I've led who helped create a lot of the data for here as well as the funders including NHGRI, and I'm happy to stop showing my screen and take any questions. As we move forward, thanks you again so much for having me and thanks everybody for listening. Well, thank you very much, Dr. Armstrong, Dr. Posey those were two excellent talks, really addressing some of the barriers as well as the opportunities and trying to address this cold prediction. I think I would ask before we have Chris come in with with some of the questions. I want to make some excellent points about where we could have, you know, modified our bold prediction as well as as maybe even been a little bit older. And, and I couldn't agree with you more about we don't need another unnecessary CDC. So, so I think where, where we'd like to go is, is that clinicians don't have barriers to ordering these things and probably with, you know, complete blood counts. There aren't enough barriers to help, we get them, we get them at times that we shouldn't but maybe Trini you could start and Jennifer you could you could comment on on how do we lower those barriers which are not just lack of knowledge or lack of acceptance but but real logistical problems in in ordering a test and getting it back and getting it into the medical record. So I'm happy to take it on. You know, I think it's absolutely critical. I mean, I do think like I, you know, hinted at Terry, you know, I think to do that. We're going to have to start to think differently and create more public private partnerships in this so tremendous amount of innovation going on in the private sector now among many of these different areas. You know, but for us to actually make this work we're going to have to I mean I don't, I doubt you want to make it operation genome speed here, but I think there is going to have to be a major effort to reach out to ask how can we actually come out the nation and these public private ways to invest in that infrastructure and make it work because as you know I think NHGRI and many other places have done heroic work trying to solve it from the ground up to try to figure out how can we create clever models to integrate into the EHR. How can we bring that from a, you know, an investigator pipeline way. But I think the reality is that for some of these infrastructure needs there's a technology innovation side, but then when it has to get implemented into the major kind of pipelines of clinical care are EMRs are reimbursement systems how we code and bill. That's going to require I think the type of public private partnership that honestly probably can only come out of a coalition that's led at the federal level. Great. Thank you. Dr. Posey. You know, I absolutely agree with what Dr. Armstrong said that, you know, as she was talking what comes to my mind that I think is a, you know, the bane of existence for a lot of clinicians, including geneticists is a lot of us spend a lot of time kind of going back and forth with insurance companies trying to get prior authorizations for testing in different forms of testing genetic testing included. And we really need to get to a point where this is considered a mainstream test, where it's understood that the type of information that a patient and their care team could get from it might actually in the long run, it, you know, improve their health reduce insurance and healthcare costs down the road. But this is actually beneficial for the insurance companies as well. So that I think that thinking also needs to change. Right now, many of us have a lot of staff members who kind of support us really just working with insurance and prior authorizations and the same is true on the diagnostic laboratory side. That's got to become much more straightforward streamlined with fewer barriers in order for us to expect that clinicians are going to have a comfort level with ordering this type of testing and knowing it's going to get done. No, excellent point. I think we've been struggling with prior authorization and and insurance challenges for quite some time. It's one of many barriers. I wonder, Dr Armstrong and the work that you're doing. Are you trying to address that and sort of lower that activation energy as well. I mean, I think what we've learned here, I'm now giving an example, Terry, of something that was done here with radiology. So it used to be and this was not my work. This is work that was led by Tim Ferriss before I got here. But it used to be that if I wanted to order an MRI for anybody that I had to go through a lot of work, a lot of effort. Mostly, I will say, you know, Dr Posing I make it sound like it's us on the phone and sometimes it's us on the phone but an enormous amount of time as other might we have an unbelievable nursing says other people carrying the water for this. And so what we were able to do this gets back to my private partnership was actually create a partnership with our local major insurers that put in clinical decision support. I feel like I'd love to show you my screen. And basically now if I order an MRI, I have to put in a bunch of criteria that to say is it a good MRI or not a good MRI, but is it red green or yellow. Does that make sense? Like should I be getting it or should I not? And I can still order it if it's red, but I have to justify it. And we actually were able to work with the insurance companies to do away with the prior authorization just based upon doing that. The problem is we can't have 27 different types of that across the country. There can't be one type here and one type there, right? We've got to be able to come together, I think at the, at your level and be able to say what is the way that we're going to do this. But insurers don't, every payer wants their patients to get the right things. But as you said, we've spent a lot of time in medicine ordering a lot of the wrong things. And so we need to be responsible about how we develop those systems and then how do we do a, you know, we've got to be partnering with the payers to figure out how we implement those systems. Great. Dr. Posey, do you want to follow up on that? Are you good? You know, I absolutely agree. I think definitely the partnership with the insurers is going to be particularly critical because I find that many times, at least for my patients, and again, this is coming from a genetics clinic event that once I escalate to a peer-to-peer conversation, many times I have a physician on the other end of the line. He works for the insurance company and they agree with me with the recommendations for testing and then they turn around and say, however, our policy is simply that we don't do testing in this type of scenario. So all my professional opinion is that I agree with you. I can't approve this. And so I think that's where really partnering and starting to think together about what are reasonable expectations and what are the guidelines going to be. And then the Dr. Armstrong said really making it truly universal so it's not a different set of guidelines for each company or each scenario. Great. Great. Thank you. Dr. Gunther, you have some questions from the Q&A? Yeah, absolutely. Thank you so much to both of our speakers for being with us today. So I'm going to follow up with one that we got in advance, which is on the same theme that Teri has been talking about the barriers to make a reasonable prediction happen. And that was, are there concerns about obtaining consent from patients to perform routine genome sequencing this routinely? What would you say about that? So maybe I'll ask you since I asked Katrina first last time. So Jennifer, maybe you want to take that first and then follow up with Katrina? Sure. So I think it's a very good question because consent is an important part of what we do when we're performing genetic testing. And something I didn't have time to get into is that it's really sort of a three-stage process. There's a pre-test counseling, pre-test discussion. What are the testing options? What do they tell you? What do they not tell you? What are the limitations of the results? Or are there types of uncertain results? What we would call variants of uncertain significance that you might have? And answering questions in that vein and also making sure people understand what might this mean for your relatives? And might you discover things about your relatives through your own testing that you may or may not want to know? And then there's really the, if we're talking about exome sequencing or genome sequencing, there really is a very well-defined consenting process. And so diagnostic laboratories will typically have a written consent that's required, unlike, say, a CDC where ideally you are getting patient consent, but that's happening kind of verbally and often very rapidly. And then at the end, the sort of post-test counseling. So we talked about your possible results. We have your results. Let's talk through now exactly what this means. And so it is a very kind of three-stage process. Part of the reason I think that we need that is simply because right now we're not in a place where we've been doing this long enough that the majority of people who come through our clinics understand the different types of genetic testing, the different types of results. I think that that's going to change over time as society becomes more savvy along with us. And so those conversations over time may be much more short, much more direct and to the point. You know what possible results you might get from your complete blood count. You might walk in kind of generally knowing what types of results you might get from a genetic test you might have in mind that you want to come talk to me about. And so I think some of that may be streamlined a little bit, but depending on the test and the potential types of results and what may or may not be reported back to the patient or what may or may not be sitting in those data over time, I do think that that discussion, that consent discussion is going to remain a critical component moving forward and people deserve to have that discussion with their ordering providers before the test is ordered and put into play. Great. Thank you. Katrina, what do you think? Yeah, I don't know. I'm probably a little more of a outlier on this one. You know, I remember the days when I trained. So one of the things I discovered is that, and Terry's being very nice not to say this, is I'm apparently the old person on the show today. That when I trained, you know, was just at the beginning of the HIV ever, maybe the height of the HIV epidemic. And I don't know how many people listening remember what we used to have to go through to get consent to get an HIV test. It was like triplicate forms, at least what I had. Nobody could get an HIV result over the phone. And so we were caring for very indigent populations. I was at Hopkins. It was almost impossible to get people tested because you couldn't be sure they could come back. And so I sometimes think that we we've got to get past the one size fits all model here for genomics and consent. And so I guess if I were going to say, I think Terry, my advice or my answer to that question would be that here is an opportunity for us to do some really great both investigation about what works when for whom. And us to develop public-private partnerships so much going on out there in the private sector around this and really come at it with an evidence base, right? Like when is it that we need to do X, Y, and Z? And how do we know that? I think we've we've consensus almost a religion and genomics. And I think it needs to move from a religion to a science. Great. Thanks. Excellent points. Go ahead, Chris. So one of the, I think, Katrina, you, I think you talked about this with your opinion leaders and networks. And then, Jennifer, I think you talked about this a little in your answer, but we did get a question from the audience. What about the role of health literacy and adoption of genetic testing, both for providers and patients? And I personally would also add genetic literacy. How do you feel we should go about increasing both of those? Okay, so shall we start with Katrina this time? Oh, sure. We can do that, Terry. We can flip up it. I think it's such a great point. I'll tell you, Chris, that one of the first studies I ever did, and this was a very discouraging moment in my career, was that we were collecting data about when we gave out BRC one in two test results. How should we do that? Like how should we do it in a way that actually people are able to use women primarily, use that information. So there are also different ways you can give risk information, right? Like you can give it as a percentage or one in a thousand, right? You guys all know this. So I was doing a bunch of studies and we decided that we would go practice these things on the pen undergrads. So I was a pen at the time and I had no real money. You know, I was living on a K award. I think Dr. Posey knows how that works. And so I was going and so we did all these surveys sitting outside of the pen library measuring their risk perceptions and communication. And do you know the amount of basic numeracy skills in the pen? These are pen undergrads. It was appalling like very few people could flip a fraction to a person, not very few, but many people could not flip a fraction to a percentage. Now, it's possible they just wanted the candy bar and they wanted, you know, me to go away. And so they weren't really paying attention. But I think the questioner is completely on the money. I think we are just massively under paying attention, not just to health literacy, but to genomic literacy. But I'm going to tell you, Chris, I think that's a job for whoever and I'm going to be embarrassed to say that I haven't kept up with all of it. But that's like a Biden infrastructure thing. We need to get this into schools. This needs to be done at the science curriculum level and this needs to be in schools and high schools. You should not be able to graduate from college. I mean, do you know that many colleges still make you be able to swim to graduate? I think it is way more important that you have basic literacy and primary care health literacy and genomic literacy than it is that you can get back and forth in a hundred yard pool. Great, Jeff, what do you say to that? Yeah, I completely agree with all of that. And, you know, I would say also in terms of the genetic and genetic literacy for more providers, you know, we've got to bust that up in our medical schools. And as part of that, so that at least the newly trained physicians coming out are ready to go and are able to think about some of these different complex concepts to have basic conversations with their patients. And of course, as always, to then refer to, you know, genetics or engage them in the patient's healthcare if there's a very specific question where they feel they need that input. But we do need to level the, or raise the baseline of a lot of our genetics education. And not only that, it really has to pivot at the same time that some of our other approaches to medical education are changing. So, you know, the way that we learned in medical school, you know, is very different from some of the approaches that are happening now. But at the same time, I think we really need some methods to train our current clinician group. So it's without a doubt, me coming out of medical school, you know, compared to what I feel like I would need to know now if I were a practicing internist, you know, that this is a field that changes rapidly. And when you think about people like my father, who is practicing internal medicine until very recently, his level of ability to engage with genetics or to think about who might be appropriate for testing was very limited because this just wasn't taught in medical schools. And it's not something that's probably as much of a component of general medicine literature as perhaps it should be, at least not in a way that it's easy for people to always kind of take that information and run with it. They have very specific guidelines, like who is appropriate to select for BRCA-1 and BRCA-2 testing based on their personal and family history. And we've got to find ways to really educate those individuals. And I think that's where some of the piloting that we're doing where we're really partnering with primary care physicians or cardiologists and endocrinologists and finding ways that we can really almost in real time talk through cases and they may be ordering the test. They may be giving back the results, but have us there as point people until they get to a place where for the majority of cases they can really discuss those findings. And then in the rare situation where there's something that really needs medical genetics evaluation, then those individuals get referred. And so I think that that's going to have to be a continuing partnership and conversation. We do talk a lot about the solution to this is sort of broad education of the public. And yet I was taught as if we were you that doctor means teacher and we should be able to explain these concepts to our patients. Hopefully we don't have to explain them to our colleagues, but some of them possibly. And are there, you know, ways that we can have little tip sheets for how to how to explain something in a way that that makes sense and yet is consistent with, you know, the facts and the evolving knowledge. Well, I mean, I will say, Tara, if I can jump in, you know that one of the we all spend some time looking for the silver linings of the COVID-19 pandemic, right. And one of the silver linings was like, look at what we did. We educated people in something overnight using technologies and new ways completely different ways. We had radiologists attending on our general medicine services with tips sheets, right. And so I think there's an opportunity and I think I've pushed on this before in genomics to really like sees this window and argue that there are ways that we can do this. It turns out the challenge and this is where you're going to need Jennifer to comment is that Terry, you and I probably like a piece of paper, right. Like that's what we're looking for maybe a phone and later I'll talk about like podcasts and tweet storms or tweet tutorials or I don't. So the promise I think for us to do this we've got to recognize that the information also has to get to people the way they want it and use it in that way. And so I do think we can do this. I think we're in a different place than we were in many of these different areas and that people are back to thinking that they don't finish residency and that's all they're ever going to know because we had a year and a half of learning a completely new disease, right. And people did it and did it better than I could ever have imagined. So I think there's an enormous opportunity here. You know, I would say in addition, in addition to the tip sheet, and really having different venues where there are these educational materials which will probably be different for, you know, patients across different age ranges. I'm thinking about my mother, you know, and her iPad that is not currently working versus my 20 year old patient who's, you know, blogging on things that you know I don't even really know. And but but also I think integrating this into part of that sort of annual preventative health care visit that you have with your primary care physician. And maybe it's part of the kind of family history and no they're not going to be taking necessarily detailed pedigrees, especially if they have, you know, 1520 minute new patient visits. But they might be able to start to drop sort of little nuggets of of information about, you know, this is something that we might want to think about it seems like multiple of your family members have acts or seems like you've, you know, there's a high risk of why. And I think even starting that early on, but when you have kids meeting with their pediatricians and they can the pediatricians kind of start to engage them in their health that they, you know, show up to their first internal medicine or their, you know, adult primary care visit, and they've already heard some of these terms they thought a little bit about this, so that it's sort of a continuous check in over time. Yeah, no, I agree. So I'm going to ask you one of the kind of challenging questions that we got from the audience, which was that they were surprised that you didn't spend a lot of time on the role of environment when you were talking about. Yeah, and they said if we present genetics as the answer to everything we can expect our clinical colleagues to continue to ignore genetics. I don't think I agree with. So I would love to hear your thoughts about that. Jennifer, you go first. Yeah, yeah. I completely understand the comment and you're absolutely right that while my talk very much focused on, you know, okay, we're good with one gene one disease but what happens if you have different types of variants that gene or what happens if you have multiple genes that may be in play. And the, the attendee who asked that particular question is right that they're the environment adds this additional level of complexity. And it's going to be in two parts. What has the particular patient in question what has their environment been up to this point is it, you know, simple or common examples to tobacco use or diet exercise, but then also how what are our recommendations for modifying that environment based on the results we obtain in the genetic testing. And of course we, we, there are certain recommendations that that we like to make in general. You know, somebody who has high cholesterol diet exercise that there are these things that I think beginning to understand how environment influences that and then how do you act on your genetic results. To change the environment and we're, we have still a lot to learn to the very pointed and very important question. And you're absolutely right that that has to be a component of those discussions and it has to be a component of our research and how, how we understand how environment and genomics interplay to impact a person's health so thank you for the question. I definitely agree and I know that everybody who's working in this field now is just absolutely focused on this holy grail of being able to figure out how do we bring in environmental exposures how do we bring in, you know what we understand from the genome how do we bring in and how we're able to modify in different ways. I guess what I always feel like when I think about this and maybe this is getting back I was thinking about one of the patients Jennifer was talking about her thinking Terry I was thinking about some of the patients I took care of like, honestly, I'll take anything I can get to help a patient. So if there's a genome that's out there, if there's some information that's going to help figure out how to help them if I can get them housing, if I can figure out how to get a treatment. So I often think we become kind of exclusionary when the reality is and we just published a piece on this before is that I think this is an artificial divide between social medicine and biology and we've always created this divide between the art and science and medicine or between social factors and biological factors and I think it's time to call it done. I think it's time to say that the person who matters is the patient and the community, and we should be using every tool we've got to help them achieve the health that they deserve and so yes to environment, yes to genomics, yes to housing, yes to all of it. I think that's where we should be. Yeah, but much, you know, much harder to do, particularly the measurement being so challenging and so confounded by many other things. I'm going to push you on that I think that's right if you decide you want to measure diet, because as you all know, like, we all lie about our diets constantly like I'm not willing to tell you what I had for lunch, I promise, particularly not on something that's being recorded. But I think, Terry, we've got a lot of great new data that are going to, you know, new ways, you know, another revolution that happened in COVID right is that we were in patients homes with all sorts of things. I think if you look at what we're able to track and understand I think it's amazing we have an investigator here who is doing this incredible study I know you guys are going to think it's crazy of trying to look at adverse childhood experiences by looking at teeth that were shed and what can we measure in teeth from children when they were shed. So I think that's the measurement issue is I think, yes, it's been hard but think about with, I mean think about the genome right. This is something we can do and we can take it on and I think we can make huge strides in it. But I would just never ask people what they had for breakfast or lunch like it's not going to work, we should just give up on it. Fair enough and I think you know one of the challenges that we get into with genetics that it's so, so easy to measure the genome now it wasn't originally but but you sort of default to that and say well look at all these cool things that we can see even though the effects are very small. So, Chris do you think we have time for one more. We have one more lightning round question which I'll ask you and do you hopefully 30 seconds each answer. What patient samples do you usually take to your genomic analysis saliva blood etc since different cells have different gene expression profiles. Do the patient samples you take differ depending on your suspected clinical diagnosis Katrina, and I would add how do you think they might differ for the school prediction but go ahead. So I would say that that is a and I had the benefit of watching some of the prior talks here and I would say that if you look at some of those prior talks, they're saying that's where we should head right so we should not only be thinking about one genome we should be thinking about many genomes many cells many analysis, at least for me we're pretty proud if we've got one genome so that's where we are today but I think we will get to tissue specific and then we will get to cell specific. We know NHGRI is going to get us there. Thanks, Jennifer. Yeah, I think I, you know, I would just add that by and large, we are operating with flood and saliva because we're trying to answer questions about what we call germline mutations or germline variants of variants that we anticipate are present in all cells of a person, but there are certain rare conditions, as well as cancer, where you might have an organ system or a part of the body that has, you know, a subset of cells where there's a different variant and you're not going to detect it, unless you've sampled that tissue or that part of the body. There are times when we are, for example, with a particular skin lesion, taking a biopsy from the skin in the lesion and then right next to the lesion in the sort of typical skin area, so we can compare and looking for genetic differences between the two. So it's a bit as you as the I think the attendee alluded to it can be determined a little bit but based on what question we're trying to ask or answer, but blood and saliva probably far and away the most common, and then others accordingly. Great. All right, well I think we want to end on time so I want to thank both of you for all the time and effort put into these talks and the stimulating discussion we've just had. And many thanks to all the attendees for the good questions that we got. Unfortunately, we can only answer a few of them, but we will, we'll try to hang on to them and keep them in mind as we try to move forward with this at NHGRI and elsewhere. I would like to remind folks that this is only the sixth. We have four more of these bold predictions we're actually going to take a month off in August. So our next one, the seventh full prediction will be Thursday, September 16th at 3pm. And that one is clinical relevance of all encountered genomic variants will be readily predictable rendering the diagnostic designation variant of uncertainty significant or the US obsolete. And that will feature talks by Heidi Rim, as Katrina mentioned of the Broad Institute and Mass General Hospital in Boston, and Doug Fowler at the University of Washington, and that will be moderated by my colleague, Ben Solomon, who is the clinical director here at NHGRI. And with that, I'll thank everyone for your attention and we'll sign off. So thank you very much.