Well, welcome to the CARDI-EP society meeting. It is a – well, I'm asked to read this out loud. So it is my pleasure to welcome you to San Diego and the Horizons 2025, the 46th annual meeting of the Horizons Society. It plays itself. All right. So if you have not already done so, please download the HRS mobile app from your preferred app store. And this is how you can participate in live Q&A during the sessions. Please scan the QR code on the screen to access this session's Q&A. And when using the mobile app, log in with your HRS credentials. Please note that the visual reproduction of Horizons 2025 either by video or still photography is strictly prohibited. So today we have a great program for you, but now in the next 10 minutes we'll have a business meeting, and I will give the podium to Glenn. Okay, thanks, Peng, and welcome, everybody. It's a pleasure to see you all here today. Okay, so let me start by talking about membership. We're really happy to see the membership is rebounding after its nadir at around the time of COVID. It's really nice to see that the trainees and fellows especially have signed up to join the society. Why is that doing that? We'll go backwards. Okay. And so you can see we're really up to, like, 300 members, which is great, getting, like, in the old days. I just want to quickly go over our financial position. We started last year with a balance of about $169,000. The meeting cost us about 20 grand, 16,000 in other expenses. We only bring in about 8,000. We don't like that number, so it jumps ahead, in revenue from the dues. And so we're losing about $29,000 each year. In the past, we used to get a lot from Medtronic and from other industry sponsors of our program. That's not going on anymore. So it's really important that all of you who come to this meeting join the society. The dues have been $60 for full members and $30 for trainees forever. We probably will increase that up to more like $100 and $50 in the coming year. It was really nice to see a lot of people have clicked through on the QR code during the poster session and lunch and signed up, so that's going to help quite a bit. We are proposing a few changes in our bylaws. I think these were available online for some of you to take a look at, and we'll take a quick vote on that. Two main points. One is that we switched the meeting from the AHA, I guess this is on auto forward, from the AHA to Heart Rhythm a number of years ago, but we never really codified that, and so we're going to vote on that. And the second is that we've added, we're proposing to add three new officers to our society. And the next two we've had, but they weren't members of our board. So the Young Investigator Awards Director and the Marketing and Communications Director. So all in favor of these changes, say aye. All opposed? Thank you. Okay, the ayes have it. So this has passed. Second, we have to have our election for the new officers. So Lee was our immediate past president, was our president and now has become the immediate past president. Patrick Boyle is proposed to be our secretary treasurer, taking over for yours truly. Nip is nominated to be the Young Investigator Award Director, and Ellie Grandy for Marketing and Social Media Director. Are there any other nominations from the floor? Hearing none, I propose this slate. All in favor? Aye. Any opposed? Okay, the ayes have it. Congratulations to our new officers. Saving time for Dan's long talk, I'm going to turn things back over to Peng. Well, that was really efficient. But before I introduce Dan, I have to say a few words about Glenn. Glenn has been the president of this society and given the Gordon Moore Lecture and has worked very hard over the past many years, serving as a secretary treasurer. And I'm just so appreciative when I got on, he was really a mentor to me and also mentoring all our council members to organize this meeting. I really thank you, Glenn, for all this very nice work. And I know that you're going to hand over to Patrick, but we always appreciate your service. Let's give him up. Thank you. So it is now my great honor and pleasure to introduce Dan Roden. Dr. Roden grew up in Montreal and received his medical degree and trained in internal medicine at McGill. He then went to Vanderbilt, where after a fellowship in clinical pharmacology and cardiology, he joined the faculty. His research focuses on the genetic determinants of abnormal heart rhythms, high-throughput functional assessment of ion channel gene variants, and the implementation of genomic and pharmacogenomic information in clinical care. After serving as a chief of division of clinical pharmacology for 12 years, he was tasked in 2006 with leading Vanderbilt's efforts in personalized medicine. Under his leadership, Vanderbilt has become internationally recognized for cutting-edge programs in this area, including a very large biobank with over 350,000 samples called BioView and the EHR-based presumptive pharmacogenetic program established in 2010. This lab has been NIH-funded since 1984, before many of you were born. He also served as a principal investigator for the Vanderbilt site of the Pharmacogenomics Research Network, and currently serves as a co-PI of the Vanderbilt site of the Electronic Medical Records and Genomics Network, the genomic-leading healthcare system, which he will serve until 2029, Network, and the Data and Research Center for the U.S. National Institute of Health's All of Us program. He has served as a primary mentor for over 60 postdoc fellows and graduate students. Five have served, or currently serve, as deans, and two currently division directors. Dr. Rodin has received many honors from the cardiology and clinical pharmacology communities, notably for us, he is a past president of the Cardiac EP Society, and has received the Distinguished Scientist Award, the Douglas Seidman Lectureship, and the Ralph Lazzara Lectureship from HRS. And from my personal experience reading through the literature throughout my career, I have read many of his articles, which are very insightful and very helpful, and he coined the term, repolarization reserve, that we use frequently on the understanding of cardiarrhythmia. So it is my great pleasure to invite Dr. Rodin to come to the podium and give us a lecture. I want to thank Pang and Glenn and anyone else who had something to do with this invitation for what I think is a really, really great honor. And I use this, and I'll go to here, and hopefully this will start, nothing relevant. Sadly, nothing relevant. I hope something magical, yes, excellent. Okay, so my assigned title was A Pathway to Precision Electrophysiology, I changed it a little bit because I think there's more than one way to do this. And what I'm going to do is I'm going to tell you a series of stories, very parochial, very sort of focused on the work that we do at Vanderbilt. But when I walked around the poster sessions and I think about the sessions that we're going to see over the next three days, there are many, many stories that all lead to this idea of precision electrophysiology. So disclosures. I did not know Gordon Moe, but I certainly heard lots and lots of stories about him. And the first Moe lecture was in 1990, and it was at the American Heart Association, so it was in November, because that's when we used to do it. And it was Brian Hoffman, whose title was The Development of New Antirhythmic Agents. And I really wish we had recorded what Dr. Hoffman said at the time, because I think it would be really interesting to go back and look. But I will tell you that the month afterwards, many people gathered in Sicily, in Taormina, for what became the Sicilian Gambit. It was a five-day working meeting, and we came up with this framework. I was fortunate enough to be there, and that was published later. Now I have to show this picture. This is a picture of many people, some of whom you'll recognize, although the ones you recognize have less brown hair and more gray hair. This is Dr. Hoffman, and this is me. So I'm really honored to be here 25 years after Brian Hoffman to sort of talk about the same kinds of issues. So I'm often asked, we have a big database of evidence, and people say you should be doing evidence-based treatment, and I'm all for that. I think that if we don't know anything more about a patient, then what we need to do is treat them with the best evidence available. And I think that most people are average for most things. So I don't think we need to spend a lot of time thinking about why they're different or why not. But everybody in this room, everybody is a little bit or far from average for some trait. And the trick is to find the people who are in the little green spaces or the little red spaces and identify them and treat them differently. And the way to do that is to start with a huge denominator. You start with a large group of people, and you try to find who is at the extremes of whatever it is you're studying. So we undertook to create a biobank to start to think about that problem. And this is the sea of samples that we collected on the very first day almost 20 years ago. So the name of the biobank is BioVU, or BioView, as Peng said, 350,000 samples. The phenotyping platform is the electronic record or a de-identified image of the electronic record. That's 4 million samples or 4 million people. And it has had a huge impact on the biomedical research culture at our place. And indeed, I think many places that have done this have noticed the same thing. Lots and lots of PIs, many, many, many users, students, residents, fellows, junior faculty, lots of publications, lots of projects. Lots and lots of NIH grants. The largest NIH grant in Vanderbilt history was to support the data and research center for all of us. I'm involved in that, as Peng said, and I won't have time to talk about it very much. But that's a topic for another time. I think the biggest impact for us, though, has been this idea of training a new generation of data scientists, people who understand how to manipulate or at least think about these big, big data sets. So when we started the biobank, my idea, at least, was to find a human trait of some kind, a disease, a drug, a drug complication, use genomic information to find where in the genome you are susceptible to that particular trait, find the genes, find specific variants that mediate that. And to do that, we have set up something called a record counter. Anybody with a Vanderbilt ID can access this, and the only question it answers is, do you have cases of X? And I could sit here, and sometimes I do this in real time. I decided not to do it in real time for this meeting, but it's a very simple drag-and-drop interface. So when I ask it, how many people do we have with atrial fibrillation, the answer is 16, almost 17,000 in the biobank, in BioView. The number for the whole electronic record is more like 50,000. And of those, there are, let me see if I can do this, there are about 100,000 that have dense array data, and the intersection is about 10,000 people. So the 10,000 people with dense array data and atrial fibrillation. So using this kind of approach, we started to do genome-wide association studies a long, long time ago. So our first adventures in genome-wide association were in, big surprise, in pharmacogenetics, and each one of these is a genome-wide association, is a Manhattan plot that has been published. So the rumor was that if you came to the lab, you had to find a pharmacogenetic trait and generate a GWAS from BioView, and that's what these are. We also started to look at electrocardiographic intervals at our own site. So the numbers are actually very small, but the signals are similar to the signals that were generated around the same time in very, very large consortia, specifically from CHARGE. And then we realized that in order to generate really, really powerful signals that could be believable, we need to contribute to other large, large efforts. So we've contributed to large efforts around the QT, large efforts around atrial fibrillation, large efforts around many, many other phenotypes. And as you can see, there are hundreds of authors on these papers because many, many people contribute. And we'll come back to the utility of genome-wide association study. I'll come back to that a little bit later in the talk. So the real change for us came about two years ago when we made an alliance with pharma partners and Illumina to sequence the biobank. We've now completed sequencing of 250,000 people. Of those, about 220,000 have gone through quality control. So we have 222,000 or so samples that have a whole genome sequence, and 11,000 of those have atrial fibrillation. So there are 11, almost 12,000 people in the biobank who have whole genome sequence. And we haven't started to think about, we've just started to think about how to use those kinds of data sets. But rare variants are going to be something that we're going to spend a lot of time thinking about. So I was in a room in 2009 or 2010 when we had this kind of idea. And then somebody said, well, we have this curated electronic health record, all kinds of diseases in all kinds of patients. So couldn't we turn this experiment on its head and say, with what human phenotype is this particular genetic variant associated? So we did that. And this is a very large study that appeared in Nature Biotechnology several years later to validate this idea of a phenome-wide association study, or a PHEWAS. And lots and lots of authors. And the reason there's lots of authors is because many of them are from Vanderbilt, but many of them also come from the eMERGE network, which we've been part of since 2007 and continue to be a part of. So I'll show you a couple of ways in which we've used phenome-wide association study in the electrophysiology space. One of them was we took this QRS signal that we got in our very, very early, very small, very underpowered genome-wide association study and asked the question, with what human phenotype does that top snip associate? And the answer was cardiac arrhythmias and atrial fibrillation and cholecystitis, which I don't think is a real signal, but maybe it is. Who knows? So we looked at this one, and Josh Denny then generated this plot. And the really interesting thing about this plot is that it's from the electronic record. So these records go back a long way. And the x-axis is 20 years. So we can generate a survival plot out of the electronic record, which is something that is hard to do with using other kinds of resources. And then I gave a talk sort of like this probably a decade or more ago at NYU. And Glenn and David Park came to me afterwards and said, wait, they have this huge data set of ETV1 variants in mice. And they were studying the role of ETV1 in development of the conduction system. And the criticism from the JCI authors, JCI reviewers was, what does that have to do with human health? And so we generated this PheWAS plot that shows that specific variants in the ETV1 locus associate with bundled branch blocks. So maybe that got them over the top. Maybe the science got them over the top. But it was fun anyway. So the rest of the talk, I want to talk about this idea that in order to personalize anything, personalized electrophysiology, personalized medicine, you have to have this dense interaction between basic science, discovery science, and then implementation science. And I'm going to give you six examples or five examples over the next 35 minutes and 47 seconds. That's what my clock says here. So the first example I want to show you is some studies that we've been doing, looking at control of the QT interval, doing this work mainly in IPS cells. And the person who's driven that is Yuko Wada, who is a very talented postdoctoral fellow, now junior faculty member at Vanderbilt. The first paper that we published, which I'll just briefly summarize, looked at these two variants, which are common, well-known variants. They're ancestry-specific. So S1103Y occurs in 8% of African-American subjects and 0.001% of European ancestry subjects. And the same for the East Asian variant, R1193Q. They both generate a late sodium current in vitro, when they studied in hex cells, or in IPS cells, for that matter. So we looked in BioVu, because they should prolong the QT, and in BioVu, there was no signal at all in a decent number of people. And then we generated IPS cells. When I say we, Yuko generated IPS cells. She generated IPS cells by taking population-controlled European ancestry subjects and putting in the variants, or taking African-American subjects who had the variant and making isogenic controls by editing that out. And it didn't matter which way you did it. There was no effect on action potential duration. So that just makes no sense at all to this audience. The action potential should be long. The QT should be long. The reason they're not is that these people, these cells are smart. And they up-regulate IKR through mechanisms that I wish I understood more, but we don't have a handle on the mechanisms yet. What was interesting, though, is that when you challenge these cells with an IKR blocker, they are exquisitely sensitive, because you take away this IKR sensitivity and you increase, you unveil the late sodium current effect. So they are very, very sensitive to generating arrhythmias, at least in vitro with IKR blocks. So I think that's part of the story with this. So that's sort of a prologue to this story. This is a patient who I've been following now for 15 years. When I first met her, she was in her mid-40s. She's congenitally deaf. She's been on beta blockers since childhood. She's taking the wrong beta blocker, probably at the wrong dose. And she doesn't want to change her medications and does not want an ICD, despite the fact her QT is 600 on this tracing. She has occasional syncope, and she says it's only when she gets hot. So we said, well, how about if you do a treadmill to just make sure that the beta blocker is the right beta blocker? And after a long discussion, she agreed to do that. And this is what happened on her treadmill. At really sort of three minutes of stage two, which is not much exercise at all. To this day, she will complain to me that she's never exercised that hard in her life. But it's not very much exercise, I will tell you that. And we made IPS cells from her. One of the first set of IPS cells we made, and they have very long action potential. So I said, this is cool. We're going to write this up, and we'll show that these cells have no IKS, and they have long action potentials. And then wild type cells have an IKS, and they have shorter action potentials. And the problem is that when you start to look at human ventricular myocytes, they don't have much IKS. And when we blocked IKS, well, so this is an old study from a Hungarian group. They blocked IKS with HMR 1556, and there's no effect on action potential duration. If you block IKR, you get a long action potential. That's not a big surprise. And the human IPS cells in our hands behave the same way. So the control cells have a very small IKS. Yuko can find it. But when we block it, it really doesn't change the action potential duration. These are examples. These are summary data. And even if you stimulate with a PKA cocktail, the action potential, the IKS gets a bit bigger, the action potential gets a bit shorter, and then you block it with HMR, and not much happens. So we've now generated, this is my patient. This is a second patient with Gervais-Lang-Nielsen syndrome. He has a defibrillator. That becomes important in a second. And these are cells that we've engineered to generate, to knock out the commonest Gervais-Lang-Nielsen variant, R518X, which does not much by itself, but when you're homozygous, you have Gervais-Lang-Nielsen syndrome. So what's the mechanism? The mechanism is that they have drastically increased calcium current, all three lines. And when you actually use a calcium channel blocker, we use diltiazem, because verapamil is messy, because it also blocks IKR. You normalize the action potential duration. And so we said, well, why don't we give verapamil, why don't we give diltiazem to one of our long QT patients? We decided to give it to the second one, because the first one doesn't have a defibrillator in case something bad happened. And this is what happens when you give a very quick bolus of diltiazem to a patient with JLN. They start off very long, they get very short, and then they get longer. So we think it's real. We have all kinds of questions about chronic therapy and whether this applies to heterozygous patients or not. And I have a whole story about that, which is for another time, but we're working on that. And the other thing that's in this MedArchive preprint that you can look at is what is the underlying mechanism. So we've done RNA sequencing on the isogenic lines, the KCNQ1 knockout that we generated, and then got a bunch of signals that looked like they were important in calcium channel regulation, and went back and knocked each one of those down individually in wild-type cells. So you knock them down, and the expectation is that you'll increase calcium current, because these are all molecules whose job in life is to inhibit calcium current. You knock them down, you increase calcium current. And of the six candidates that came out of the RNA sequencing pool, three of them looked like they increased calcium current, and of the three, this one, RRAD, looks like it also increases action potential duration. So that's an interesting story. And RRAD has a whole story from people like Steve Marks and Jeff Pitt looking at RRAD as a regulator of the adrenergic response of calcium current in mice and other settings. We also have knocked down KCNQ1 by itself, and we see the same effect. That's another line that we're pursuing. So I could talk about this forever. I'm not going to. So the next story I want to tell you is a story that is more clinical, and it's the Atrial Fibrillation Precision Medicine Clinic that Ben Shoemaker now runs at our place. So my interaction with this story starts with this 29-year-old man who shows up at my clinic because he wants to get tested for hypertrophic cardiomyopathy. He has a very strong family history, which you can see on the slide, and he knows what the mutation in his family is. It's an MYH7 mutation. His cousin is the proband. He comes with a transthoracic echo, which is completely normal, and his ECG is normal. I said, look, it's the year 2019 or 2020 or whatever it was. We can test this variant. It's $49 at our place, so it's a cheap test to do, and you'll be negative, but because we're in genomic times, we're happy to do it. And then four weeks later, I look at my clinic schedule, and he's on my clinic schedule, and I talk to my nurse, and I say, well, that must be a scheduling error because people make appointments all the time, and they forget to cancel them, and that's what happened here. And she said, no, no, no. I'm telling you, he really needs to see you. So, number one, he is mutation positive. Number two, he's been hospitalized twice for atrial fibrillation, and his MRI is normal, and over the course of the last five or so years, he's been managed like any other patient with atrial fibrillation, with rare episodes, controlled with propafinone, and there's no evidence that the mutation affects his ventricle. So I think this is still a MYH7 mutation that has something to do with his atrial fibrillation. And the reason I think that is because Ben and Zack Yaneda, who at the time was a fellow, now a faculty member in our EP group, did a study in the top med resource, so 1,300 or so patients with early-onset atrial fibrillation, AF onset before age 60, and when you do whole-genome sequencing on them and then take the whole-genome sequence data and ask the question, are there pathogenic or likely pathogenic variants in genes that might make sense? That's about 170, mainly channelopathy and cardiomyopathy genes. The yield was about 10%. And the younger you are, the more likely it is you have a variant, and the variants that we saw, the most common was Titan, which that association is known, and the second commonest was MYH7, and then a smattering of other things, including a channelopathy gene occasionally. So Ben, with this, and our Department of Biomedical Informatics set up a best-practice alert saying, your patient is young, they have atrial fibrillation, doesn't look like there's any other good reason for them to have it, think about whether you want to refer them to the Atrial Fibrillation Precision Medicine Clinic for evaluation. And part of this is supported by an R01 that Ben has to study this problem. So these are the preliminary data. They were reported at this meeting last year, and there's a manuscript that now exists in MedArchive, so you can look at that as well. So 250 or so patients who had clinical genetic testing. They were positive in 48, 20%. So the people who refer to our clinic have a nose for who's going to be genetic. I think that's their explanation for that. And 195 are negative. Of the positive ones, about half had their therapy changed in some way because of the result of the genetic test, and that's a really important finding. And I hope that over the course of the next decade or two, a lot more will have their therapy changed by things like gene therapy, which we're going to hear about a little later in this afternoon's session. So I'm not going to say anything more about that, but that's one of the things we think about a lot as well. Among the negative people, no rare variants in 20%, a couple of heterozygous carriers, and then lots of variants of uncertain significance. Of the variants of uncertain significance, there are some that look like they should be pathogenic, but they don't have enough evidence to be pathogenic yet. So those are what we call suspicious. So the next thing I want to talk about is the variants of uncertain significance problem, because this is a huge problem, as I'll show you, in developing the idea of using genomics as a routine in clinical care. So this is a story that my colleague Prince Ken and Carol took care of. This is a 7-year-old with a witness cardiac arrest on the first day of school. First rhythm is VF, and he was resuscitated by two AED shocks. That's a miracle all by itself, of which this community should be extraordinarily proud. No family history. This is his ECG. Long QT, and when he exercises, he gets bigemini. So he could have a CPVT, he could have long QT, he could have something like that. We do genetic testing in-house, and he has two variants in a gene that makes some sense, calmodulin 2, but those have never been seen before, so they're classified as variants of uncertain significance, which means you can't start to use them in clinical care. You can't start to use them to test a family, for example. So the American College of Medical Genetics and Genomics, ACMG, has come up with criteria. This is the original article. They've been revised a little bit since, and there are currently five classifications, benign or likely benign, pathogenic or likely pathogenic, and then VUSs in the middle. And the criteria are complicated, but basically you can look at large population data sets like the UK Biobank or the All of Us resource or our BioView, and you can ask the question, are they very common, are they rare? Because common, common variants shouldn't be pathogenic, despite the S1103Y story. And do they segregate with disease? Do lots of people with a variant have a disease, lots of people without the variant have a disease, in a family or in a population? We're also very interested in using computational predictors. The latest kid on the block is Alpha Missense. It doesn't seem to perform much better than some of the other really excellent predictors, and eventually the predictors will be perfect and we won't need to do real experiments, but not for a while. And what I want to talk about is this idea of using functional data, developing functional data to deal with the VUS problem. So how big is the VUS problem? Andrew Glazer, who was a postdoctoral fellow, now a junior faculty member at Vanderbilt, did this study looking at 50 clingent-affirmed cardiovascular disease genes, so the channelopathy, cardiomyopathy, lipid genes. And in 2018, most variants were VUSs. Some of them are pathogenic, some of them are benign, and some of them are conflicting. These are annotations in the NIH resource called ClinVar. And in 2025, the PLPs and the BLBs haven't really changed much, but the VUSs have gone drastically up. And that's because we're doing a lot of genetic testing in patients like the ones I've shown you, and usually we find something that's never been seen before, so that's why we call them VUSs. The problem is actually worse than you think because that's what's in ClinVar, but about half the variants that we ever see in our large biobanks that we can access aren't even in ClinVar. So they're just there. They're usually one or two examples of them, and that's where we are. So most of these sort of cardiovascular disease genes, a smattering of known pathogenic, a smattering of known benign, a smattering of conflicting, and then most of them are either VUSs or not even annotated yet in large data sets. And that was Megan Lancaster, who's also a junior faculty, also a former fellow, now a junior faculty. So Andrew and Brett Kronke, who's another former fellow, now faculty, looked at SCN5A six or seven years ago, and at the time they looked at it, there were 1,400 variants that were reported, and most of those are VUSs, a small number of pathogenic, likely pathogenic, and small number of benign. Now, I work with this man, Fritz Roth, and I'll show you his picture in a little bit, and Fritz did what he calls a back-of-the-envelope calculation. So take the number of people on the face of the planet, take the size of the genome, and the mutation rate from generation to generation, and you can calculate, or at least Fritz says he can calculate, that every SNP that doesn't kill you when you're an embryo should be present in around 50 people on the face of the planet. Now, that's... and why haven't we seen them all? We haven't sequenced everybody in Uganda or in Chile or in Papua New Guinea. So I think the diversity of the human... Am I allowed to use that word? Sorry. Yeah, the diversity of the human genome is something that the more we know about it, the more it benefits all of us. I'll just... that's a little political statement. So the problem isn't this. The problem isn't these 1,300 variants that Brett and Andrew curated. The problem is that there are going to be 38,000 variants in the cardiac sodium channel gene, SCN5A, and so we can't be doing functional studies on them one at a time. So we have two ways that we've thought about this. One way is robotic, high-throughput robotic patch clamping. I'll give you one guess what year this picture was taken, and this is standing in front of the synchropatch machine, and this is the kind of output that we get, and then we're going to hear from Andy a little bit later, but Andy and Andrew have been working together, both in Australia and at Vanderbilt, to validate this system, and they have really interesting data, which I'm sure he'll talk about. I don't know why my slides are advancing by themselves, but they are. So the other way to do it is to say, let's take a region of a gene and make all possible SNPs in that region. So we take the cardiac sodium channel gene, for example, or a piece of the cardiac sodium channel gene, make every single possible variant, and then stick them into cells. So you have millions and millions of cells, each one of which is expressing one cardiac sodium channel that is either wild type or variant, and some of them have many, many representations, some of them only have a few. The other way to do this is to do in situ editing and something like cardiomyocytes from iPS cells. It's a little more cumbersome and not quite as high-throughput, but has advantages as well. So then you have a big pool of cells, and you take that pool of cells and do something to them. You count how many channels get to the cell surface or not, how many of them survive a drug challenge or not. Those are important details which I'm not going to talk about. And then you do next-generation sequencing on the pre- and post-exposure pools, and you do some fancy arithmetic, as Andrew says, and you come up with what we call a variant effect map. So the variant effect map is amino acid position on the X axis and every single possible variant on the Y axis. This is what a variant effect map looks like. So this happens to be a protein that's... I can't remember exactly how long it is. 149 amino acids, I think. I have to put my glasses on. And then down on the Y axis is every single possible amino acid substitution here. And each cell is color-coded. Blue is benign, green is bad. And dark green is worse than light green. This happens to be a calmodulin map. And so we use that map to actually assign pathogenicity to the two variants that we saw in that 11-year-old, and indeed both of them are pathogenic. These are in cis. We now know that. And at least one of them is de novo. So we and a group of others around the country now have a pretty large NIH-supported grant. At least today it's NIH-supported. I don't know it'll be supported tomorrow. And we are trying to do all the big Mendelian disease genes. So a group at Vanderbilt, Fritz Roth at Pittsburgh, a bunch of people at Stanford, Ewan Ashley and Vicky Parikh and then Callum McCray at the Brigham. And these are some of the maps that we've generated. I could talk about these forever and ever and ever. So this is 22,000 variants. This is about 20,000 variants. This is 13,000 variants. This is about 18,000 variants. And a dense, dense map. And I'll just highlight for you, this is domain 3 of the sodium channel. And you'll notice that there's the 3,4-linker right there. And most of the variants, the red ones are loss of function, but most of the variants in the 3,4-linker are gain of function. Not a surprise, but we can now annotate every single one of them. And then there's other kinds of biology that I wish I had time to talk about, like that end of KCNQ1, no matter what you do to it, you increase cell surface expression. Increased cell surface expression, we think, is that this is a ubiquitination signal, and anything you do to it decreases the retention of the channel in the cell. So not only will these be useful for clinical purposes, but they turn out to be really, really useful to understand the underlying biology of these large and important proteins. So I started life as a pharmacogeneticist, and I still have aspirations to do that. That's basic science and clinical implementation. The very first drug that I got involved with, with Ray Woosley as my mentor, as a fellow, was Enconide. Enconide is like Fleconide, only it has more complicated pharmacogenetics and pharmacokinetics. So in our first study of Enconide, we did a lot of things. One of them was to give each of 11 subjects that we were studying with high-frequency PVCs, we gave them each a 25-milligram dose and followed plasma concentrations over time, and then related those pharmacokinetic data to the outcome of chronic therapy. So in this group of 10 patients, the elimination was rapid, the plasma concentrations were extraordinarily variable, and there was a marked pharmacologic effect, which we thought was a good thing at the time, not so sure now, but there it is. And then there was an 11th subject who got to very, very high doses with no pharmacologic effect and very slow drug elimination. So it took us a little while to figure out what that was all about, and it turns out that Enconide is a prodrug, needs to be bioactivated, just like clopidogrel needs to be bioactivated, codeine needs to be bioactivated, and it's bioactivated by this specific enzyme, CYP2D6. And CYP2D6 is polymorphically distributed in the normal human population, at least in European and African ancestry subjects. So this is the kind of cartoon that you get when you look at CYP2D6 or genes like it, like CYP2C19. There's a small group of people who, in this case, are homozygous or compound heterozygous for loss of function variants, and they are poor metabolizers, so they can't bioactivate Enconide. That was my 11th patient. There's a group of people who have functional gene duplication, so they're ultra-rapid metabolizers. That can be a problem for drugs like codeine. And then there's a group of people who have either no or one variant, sometimes you can distinguish those two, sometimes you can't. In this cartoon, there's two antimodes. So based on that kind of thinking and drugs that have this kind of Enconide-like or clopidogrel-like story, you would say, well, implementing preemptive pharmacogenetic testing should be like the lowest-hanging fruit of all of pharmacogenetics. There are common alleles. They have big effect sizes. There's no anxiety about telling people they're susceptible to cancer or to heart disease, this is just drug response. And the electronic record systems can do it. And so we have a big program called PREDICT, where we've been doing preemptive pharmacogenetic testing for about 15 years. We started with clopidogrel and CYP2C19 homozygous. We're now up to 25 drug-gene pairs. We have clinical decision support that's real-time, so if a prescriber prescribes a drug and the system knows about the genetics, it immediately tells him or her the genetics are not what you want and consider another drug. Over 20,000 patients to date. And it's not as simple as we would think. And, again, I would talk about this forever, but I think there's other things to think about. So this is the cartoon that everybody in pharmacogenetics likes to think about. But the fact of the matter is that there's variable responses to every drug that we ever use. And the variable responses are not generally driven by single high-effect size variants. They're driven by multiple variants across multiple genes. So at this end of the distribution, somebody might be homozygous at each one of the loci that controls response to this drug, and at this end, they're homozygous for the other set of variants that drive high responses. And in the middle, there's a mix of people. So the idea is that the genetic background controls how people respond to drugs and how people respond to diseases and how people respond to rare variants that drive disease. So I showed you this before. This is a genome-wide association study and around 100,000 people looking at variability in the normal QT interval. And there are multiple hits across multiple genes. The top hit is in NOS1AP with a p-value of 10 to the minus whatever you want, 10 to the minus 200th. The effect size is tiny, 3.5 milliseconds is something that the human eye cannot detect on a normal 12 lead ECG. So based on these kinds of data, we and many, many others in the field 10, 15 years ago to look at the effect of individual SNPs on your favorite phenotype of interest, susceptibility to atrial fibrillation, drug response, modulation of the clinical phenotype in congenital diseases. And there's lots and lots of papers and I think most of them are probably only scratching the surface. I call this SNPology. Single SNPs don't modulate, don't have large effect sizes to modulate those common phenotypes. I think we've learned that. So the idea is to combine all of those SNPs into what's now called a polygenic risk score. So you put together lots and lots of genetic variants. I was a math guy in college and so this is not really very complicated. You take each one of the loci in a genome-wide association study for QT, for example, and you say, well, do they have zero, one, or two of the variants that confer a slightly longer QT than normal? And then you multiply it by the effect size, which for the NOS1AP variant or variants was 3.5 milliseconds, and you come up with a score. So what do we want from a polygenic risk score? If you do a case control study, for example, you want to find that the cases have a higher burden of loss-of-function variants or high risk alleles and the controls have a lower burden of high risk alleles or a higher burden of low risk alleles. So that's what we want to see. And that's a static version in time. You look at people once in their lifetime. But what you can do in something like Biobank or in other large resources is you can look over time. And that's really a more interesting and important way of looking at it. So these are atrial fibrillation data from the UK Biobank published last year in PLOS1. And the idea is that when the sperm meets the egg, you can do a polygenic risk score and tell that person you're at risk for Alzheimer's disease or you're at risk for atrial fibrillation when you're 75 years old. Well, nobody wants to do that because over time is when the risk develops. And what's really interesting, I think, about the polygenic risk scores is not whether they predict anything but how much more the high prediction looks than the low prediction. It's the difference between the red line and the blue line that actually gives you a sense of what the value of a polygenic risk score might be. So sometimes there's not much spread, and sometimes there's lots of spread. So this is really pretty interesting. The group at the UK, atrial fibrillation was one of 30 phenotypes that they looked at in that article. And you can't read all the phenotypes, and that's why this one is atrial fibrillation. But there are lots and lots of phenotypes that behave the same way. Not much prostate cancer, big, big difference. So those are the kinds of things that I think polygenic risk scores are going to be with us pretty soon in clinical practice. We did a study looking at cases of TORSAD and controls. This was with a big collection that we had through a LeDuc Foundation effort and in collaboration with Elijah Bear, who you're going to hear from later today. And it looks like the case control picture that I showed you before. The cases are distributed a little bit toward the right. And the p-value, which nominally for this comparison needs to be 0.05. So less than 10 to the minus 7th we were pretty proud of. Connie and Arthur and Rafik and many others in Amsterdam have done the same thing with a polygenic risk score for baseline PR or QRS, predicting change in those indices with an azomalene challenge. And just to stray away from EP for a second, if you develop a polygenic risk score for coronary disease, that predicts response to PCSK9 inhibitors, which are pretty expensive drugs. And if I ruled the insurance world, I would say, fine, I'll give you your PCSK9 inhibitor as long as you can show me that you're at risk through a coronary artery disease polygenic risk predictor. Because if you're at low risk, then there's no value in using that expensive drug. So another story around the same idea. We were part of an international consortium that looked at KCNE1 variants. So KCNE1 is the disease gene for so-called type 5 long QT syndrome. There were 26 sites around the world, 89 probands with pathogenic variants. And 63 of those had D76N, which is the originally described variant in this disease. And nine of them are in our clinic. Nine individual probands. So the obvious question is whether these are related to each other. And the way you can do that is you can do something called identity by descent. So this is a cartoon that shows the idea of identity by descent. Here's a founder who has a deadly mutation. And this is the region of the gene where the mutation occurs, and the mutation is right at that little line there. They mate with someone who is normal, and then they generate offspring. And at each time there's the mating, there's a crossover event. So not all of the chromosomes are inherited in this fashion, but there's a crossover here, and there's a crossover here and here. So this person doesn't get the mutation because the region that they inherit from mom doesn't have the mutation. These two do. And over the course of generations, the people who have the disease have a smaller and smaller portion of shared chromosomal information. So after many generations, you can see this person is a mutation carrier, this person is a mutation carrier, this person. And there's shared segments. And the larger the shared segment, the closer they are related to each other. So Piper Below, who is now director of our Genomics Institute, Genetics Institute, and Megan Lancaster, whose picture I showed you before, looked at our D76N probands, and they share two to 60 centimorgans. That's two to 60 million base pairs. So that's a small segment if you're a genomicist. For us, it's a huge segment. But that's the region that they share, and they're all probably descended from a common founder ancestor that came to the Tennessee area around 150 years ago. So that's an interesting story. Piper is particularly interested in identity by descent and using that for discovery. And so she has gone to BioVu. This is old data, because this is when we only had arrays, and we had arrays in about 100,000 people. And she separately analyzed European, African, and Hispanic ancestry individuals. So in the European ancestry set, there are 3,500 people who are first-degree relatives, so siblings or parent-child. And then the more distant the relationships, the larger the number. So by the time you get to sixth-degree relationships, which is about second cousins, there are 144,000 pairs who are second cousins, many of whom know about it, some of whom don't. So using that kind of information, we went in and asked BioVu, are there people in the biobank who share this chromosomal segment? And the answer was there are, in addition to the nine or 10 mutation carriers, we actually have 10 proband mutation carriers now, there are about 22 who share the same chromosomal segment in BioVu. They're not the same individuals. So we have ways of making sure that we don't double count. And I'll show you their QT intervals in a second, but I'll just show you, this is the distribution of normal QT intervals in our biobank, in our electronic health record. These are patients who have no confounding medications and no heart disease. And so that's the distribution of 30,000 normal QT intervals done in 2011. And these are the data that Megan generated, either from the clinic or from BioVu. You can see that the QT intervals are widely distributed and mostly at or above the upper lunar normal, but some of them are clearly within normal limits. So the question was, why is there such a big distribution? And the answer is that at least part of that variability is explained by the polygenic risk score for QT itself. So there's this idea of an interaction between rare variants and common variants that capture the genetic background. So Brett Kronke, again, whose picture I showed you before, has started to do this experiment where we're trying to figure out what it is about a polygenic risk score that actually tells us something about the underlying biology. So he's taken two people, one of whom has a polygenic risk score of 1% for QT, and the other one of whom has a polygenic risk score of 99% for QT. And he's done a couple of things with them. One is that he's compared baseline action potential duration, or these are actually extracellular field potentials that you can do with the iPS cells, and there's not much difference at baseline, which is sort of what we expect, because the QT interval in normal people varies, but it doesn't vary a huge amount. But if you then expose them to E4031, an IKR blocker, the ones with the high score get very, very long. The ones with the low score don't get very long at all. And what's really nice about this experiment is that you can actually then do proteomics or transcriptomics on the high and the low polygenic risk scores cells and start to come up with molecular pathways that mediate that polygenic risk score argument. The polygenic risk score is a statistical genetics argument, and we need to understand the underlying biology. And this is the last data slide. So he's also inserted a known pathogenic variant into these cells. And what's very cool is that these are the action potentials in the 1 percentile cells, and these are the action potentials in the 99 percentile cells. So we really think that there's an interaction and there's an opportunity to understand the underlying biology that modulates that interaction. I don't think it's going to be a single SNP, though. So I've talked a little bit. My topic was pathways to precision electrophysiology, starting with cases in some cases, starting to probe the underlying biology in many, many ways. And I've been parochial and talked about our own work, but we're going to hear this afternoon about many other works, and this is the same idea. And then thinking about how to implement that in clinical practice. And I think we're getting there for at least some of the examples that I've shown you here. And I want to emphasize again that 25 years ago, when Dr. Hoffman talked and the Sicilian gambit came out, the focus was then on arrhythmogenic mechanisms. I don't think that part has changed very much. So I'll just close by showing pictures of people with whom I work, all these pictures. The rule of this slide is that if you leave Vanderbilt, your picture disappears. So these are the people who are currently still at Vanderbilt, with whom I work on a daily or weekly basis, and I'll just emphasize that we have this new network where we're going to try to implement some of this genomic information. So I thank you again for the honor of doing this talk, and happy to take questions if there's time. Thank you. Thank you so much for this very exciting talk. I would like to invite some of the trainees, if possible, to ask the first questions. Anybody here wants to ask a question, please? No trainees. It's brave of you to do that. That's a challenge. So, oh, somebody's coming. No? No. Everybody's leaving. That's right. I drove them away. I drove them away. Anybody else have any questions? There's a microphone over there. Please feel free to use it then. In terms of clinical applications, oh, go ahead. Hi, Dan. Nice talk. I'm Andy from Australia. So with your comodulin variants, we have three comodulin genes, one, two, three, they're identical. So do you have any comment on the other two? The comodulin story, well, yeah, I skipped over that. In the interest of time, so the comodulin is really interesting because it's a short protein, 149 amino acids, and it's encoded by three separate genes, each one of which encodes exactly the same protein sequence. So presumably you have three different genes because the regulatory regions of those genes will tell them where to be expressed or when to be expressed or when to not express. So the comodulin story is interesting. The map, the variant effect map is the same for all three of them. So that's a comodulin variant effect map, and it doesn't matter whether it's com, one, two, or three, we think. At least, I should say, Fritz thinks. Hi, Dan. Excuse me. Two questions, if that's okay. Brief one on D2, the D76N. Did that ancestor come from Ireland originally? Oh, I have no idea. I mean, you know, so, you know, if you're in Iceland, you can track where your ancestors came from and you can go to the churches and what have you. We don't know the answer. Well, you may remember from that paper that over 50 percent of the rest were contributed from Northern Ireland. Right, well. So I was just wondering if we join them together at some point. It's entirely possible, and the answer is I don't know the answer. It's possible anyway. Yeah. The other thing was you've obviously mentioned the, or it was mentioned by Peng, your utilization of pharmacogenomics early on in the electronic health record. Are you using it yet in terms of at least polygenic risk or for predicting risk of TORSAD in your patients? No. No. I mean, so the, one of the things that I've learned over the course of trying to do this is getting an institution to actually sort of implement something means you talk to the clinical pathologist and the clinical pathologist will say, well, what's the assay and how do we validate that assay and where are the positive controls and where are the negative controls and how do we stay CLIA compliant and how do we stay CLIA-CAP compliant? And so all of those things are obstacles to developing clinical implementation of polygenic risk scores on a population scale. Some companies are offering that now, but that's sort of small scale, I think, and we would like to see it large scale, not just in the electrophysiology world, but across cardiovascular science and across all of clinical medicine. And so how to get that done is actually going to be a challenge, but we've been at least thinking about it for about the last half decade because we have an institutional culture that values moving to precision medicine and we have pathologists now who want to do it. It's just a matter of figuring out what the pathway is from here to there. And I think particularly when a lot of the risk may be driven by rare variation, that is always going to need that polygenic risk score to interpret the risk. Well, I think the polygenic risk score sort of captures what are called the genetic background. I think that's, you know, that's sort of our version of SNPology, it's not one SNP, but millions of SNPs. And I think the one way to look at those, that last two sets of action potentials is that something in the genetic background of the high polygenic risk drives arithmetogenic behavior in the setting of a rare variant. The other way that has been pointed out to me by my wife is that something in the genetic background of the one percentile people is protecting them. And I think that may be a more rational way of looking at it, actually. So it would, and the polygenic risk scores are a statistical trick. I mean, it's what, that's, you know, Peter Donnelly is a statistical geneticist, he's not a clinician. And so they can do that statistical trick, but what the underlying biology is, what is it that drives susceptibility or not in the presence of a rare variant, in the presence of a drug, or in the, just in the presence of being a person, is something that I think we still have to get a handle on. And this is an opportunity. But it may still not prevent it being useful, though, clinically. Oh, I think it's part of the future. Okay, thank you. Okay. Dan, in terms of clinical translation, how about the CYP2D6 and so forth? How come we can, physicians still cannot click a button on medical records? I get a, currently I get a phone call or an email, I got an email from one of my young colleagues three days ago saying, tell me again how to order the pharmacogenetics test. The problem is, it's in Epic, it's in our instance of Epic, and you can order it. And we've managed to figure out how to pay for it. But the problem that we have is that the time between ordering it and actually having the data is about a week. And sometimes that's okay, and sometimes that's way too long. So there are point-of-care tests, for example, for CYP2C19, for clopidogrel, that you can do in 45 minutes. But that's one-off, and it helps you take care of that patient at that time, but it doesn't do anything sort of long-term, and it doesn't sort of bend the curve in terms of other drugs that they might use. So we still believe in this, but we've had discussions with our pathology leadership now of whether we need to have a parallel track to introduce those kinds of variants early on as well. The perfect vision is that when you turn 40, you have a pharmacogenetic panel that then either sits in your electronic record, or better yet, is accessible by a QR code on your phone, so when you go to your next hospital, you can actually tell them what your variants are, and then they can figure out whether you should get clopidogrel, or whether you should get 5-FU, or whether you should get propafedone, yes or no. So those are the kinds of things that we'd like to do, but it's the mechanics of implementation are hard. So we do it at our place, and it works when it can work. Are there any questions from the panel? No? Okay. Thank you so much, Dan. I was wondering, in your map there, there are the genetic information, if we were to think about some of the epigenetic, whether that would... We would love to think about how that works. The way we look at the variant effect maps now is that you make the library, and you study that library once, and then you expose the library to a drug, and you study it a second time, and you put the library into a cell line that has another variant in it, and that's a third study, so we could do that over and over again. And whether we'll ever get to the point of being able to do epigenetics on those libraries or not, I don't actually know. We're writing the competing renewal now, so stay tuned. Any other questions? Okay. Well, Dan, I want to thank you very much again for giving this very exciting talk to lead off this afternoon. We have a very modest honorarium for you, but a real plaque will be coming to you email.

CARDI-EP Society Meeting

Horizons 2025

membership dynamics

financial struggles

bylaws amendments

precision electrophysiology

genomic research

personalized medicine

Atrial Fibrillation

pharmacogenomics