Hello and thanks everyone for joining us today. My name is Jessica Profato and I'll be your moderator. I'm a genetic counselor and a senior manager of product management here at Ambre Genetics. I'm excited to be joined today by two of our expert variant scientists, Tina Pessaron and Marcy Richardson. Tina received her master's in microbiology, immunology, and molecular genetics from UCLA, then received her master's in genetic counseling from California State University, Northridge. Prior to joining Ambre Genetics in 2011, she worked as a clinical cancer genetic counselor for eight years. She is now the director of Ambre Genetics variant assessment program. Thanks for joining us today, Tina. Thank you for having me. Dr. Marcy Richardson graduated from Georgetown University with her PhD in tumor biology in 2009. She then joined the University of California San Diego as a postdoctoral fellow in epigenetics. She began working as a variant assessment scientist at Ambre Genetics in 2015. We're glad to have you here today, Marcy. Thanks, I'm glad to be here. So as you all know, there's a lot of discussion in the field of genetic testing and genetics as a whole regarding the best approach to ensuring the most accurate and informed analysis of genetic variants so that we can report clear meaningful results to genetic counselors, physicians, and ultimately patients. Today we're going to spend this time addressing observations from our team as it pertains to variant assessment and ways that we're really driving the field forward for both quality and access. So to kick us off, you know, there's a lot of new studies happening all the time that inform improvements to variant assessment. Tina, I would love to start by hearing from you about some recent improvements in variant assessment and really the potential benefits of using these studies and innovative approaches. Yeah, no, that's a great question. You're right, you know, there's definitely a lot of studies being done and data being published that we can leverage in improving the variant assessment process. You know, the original framework that ACMG provided for the community is really broad categories of evidence types and not a lot of, you know, nuances that are defined, but they recognize that, you know, there would definitely need to be improvement in the rules over time and as data is collected and we better understand these, you know, genes and diseases that we're working on, we can improve that process and as well as the variant types. So, you know, one example I can think of, for example, more recently in the last couple of years, you know, we were working on better defining the framework for how we weigh variant types that commonly are defined as pathogenic, like nonsense, frame shifts, and gross deletions and duplications. These all fall under the, what we call the PBS-1 code, and in this code, in this evidence type, you know, the way that ACMG provided it, it was just a very strong line of evidence, but those of us who obviously look at these variants every day recognize that it's not that simple, that there, you know, are considerations that need to be made in order to give it the full weight. So, we were working on this framework to help better define what questions need to be asked and when you get the full weight versus when you don't get the full weight of the evidence. And one of the ones that was tricky was the gross duplications. These variants typically, historically, were thought, were classified as BUS mostly because we didn't know where in the genome they were going to, they were deposited. So, the test that we do, the multiplex type test that we do, or the array test that we do, doesn't tell us where it's inserted, we just know that there's an extra copy there. So, it was really important to know where that is before you can make any assumptions on whether it's disrupting the gene or the protein. So, we did a study. We said, let's go back and let's look at all these variants that we've identified so far. It required that we do additional genetic studies to find the location, and it's a study that we published actually, Marcy is the first author on that, and it showed that almost always they were in tandem. They were within the gene and would potentially impact the protein function. So, because of that, we were able to then go back and say most duplications are going to be in tandem, and so we do have a high prior probability that these are going to disrupt the function of the protein, and we can weigh them now appropriately, now that we know this information, and that was incorporated into the PBS-1 guidelines, and that is now something that the majority of the groups have now implemented. So, having that data was really important. There are other examples looking at evidence types. So, I think of this as two different categories. So, the one I just talked about helps us improve overall rules. So, we have a general framework. We need data to then make that framework better, but what about better understanding evidence types? So, this is a second category where, you know, if you have enough data, then you can make a decision as to how much evidence towards pathogenicity that provides. So, you know, looking at, for example, in the mismatched repair genes, a common data point is the tumor data, whether we have microsatellite instability or loss of protein on immunohistochemistry, but we didn't, you know, how much weight do you give that given that there are also sporadic cases that can look similar? So, we've done the analysis and, you know, looking at large numbers of tumors and was able to then come up with what is the evidence path likelihood of pathogenicity for these types of tumors. So, you know, definitely with each evidence type, the more data you have, the more information you have about how much you can weigh that beyond sort of just what was defined by the original ACMG. And we're trying to, you know, also look at things on a global level and looking at protein structure to help us understand these things. So, you know, obviously a lot, the more of these studies that are published, the better our rules are going to be in the better our data points are going to be. Absolutely. Thanks, Tina, for walking through those great examples. And, Marcie, what about you? Can you share some examples of studies, you know, that you think have really been impactful in driving variant assessment forward? Oh, yeah. Tina, I mean, Tina hit on a bunch of them, but I think, you know, one area that's really exploding right now and super important to talk about is functional studies. Very recently, the Sequence Variant Interpretation Group, also known as the SVI, came out with a nice publication that helps us further understand how to empirically weight functional studies. And this has led to a huge improvement for data-driven application of ACMG rules. So this would include not only how to weight sort of the historical and gold standard functional assays. So, for example, at ACMG this year, we have a publication looking at a gold standard assay for BRCA2, and the impact that these new guidelines had on BUS reclassification rates was well over 80%. So now that we have those tools in our toolbox, hopefully we can approach the application of functional studies in a more uniform way. So that's been a really big advancement in that area. And then also these really huge multiplex assays. So they're called MAVE assays, which is an acronym, Multiplex Assays for Variant Effects. So these are some of the newer ultra-high throughput functional studies that basically look at every single nucleotide variant that can be made, whether it was observed in a patient or not. And they can really impact the interpretation of novel variants or very rare variants, and can make major advances in that way too. And then, of course, RNA studies, right? People sometimes forget that an RNA study is actually a functional study. And we're just now embarking on this new era of understanding more and more about RNA. So it's a really challenging topic, and I think the fact that we are now offering a paired DNA RNA test has made huge advances in understanding this field. We can, basically, we can look at variants that we expect to have an RNA defect, such as those ones at intron-exon boundaries. But we're also solving mysteries by identifying these deep intronic variants. So APC is one of my favorite examples because it has a clear phenotype. And, you know, like 25% of people with polyposes don't have a genetic explanation for their disease. But we keep finding deep intronic alterations in APC that do explain them. So these families are starting to get answers now. And even beyond just, you know, individual people and their variants, we're learning more about, you know, exon level expression data, gene level expression data. And we can use all of this information to influence things like in silico tools. So, you know, we can better understand how to set thresholds for those. So knowing more about RNA is helping us in many, many different ways. Absolutely. All great points, Marcy. It's been really exciting to see how functional studies and the inclusion of RNA analysis is really helping to not only drive down variance, abundance, significance, but like you said, also find answers for more patients. So shifting gears a bit here, you know, I do want to dive into another hot topic in the field, which is really artificial intelligence or AI for variant assessment. AI has definitely had an impact on our industry and will continue to have an impact. What are your thoughts on relying really on AI alone for variant interpretation? Yeah, so I think that's, you know, another hot topic, obviously, with the advancement of computers and computational biology, this is, you know, definitely something, an area that can help the process of variant assessment. You know, these are tools that people are creating, and more and more of them are coming out. You know, I think we do need to be careful how, you know, how it's being used. The tools are as good as the data that goes into it and as smart as the humans who created it, right? So, you know, we're still, you know, at the forefront here, and we're, you know, working towards improving all of these tools that are coming out. You know, they essentially are using the frameworks that we've developed and, you know, gathering data and trying to better understand and sort of help the end users in making decisions similar to what you would do manually. I would say that, you know, again, we do need to be cautious about how we use these tools. They are, you know, as we've talked already, there's always nuances. There's, you know, exceptions to every rule when it comes to biology. So, you know, we're, as humans, we are trained to sort of look for that. We're looking at, you know, the clinical data. One of the things I think that, at this point, is lacking for a lot of the tools is that clinical correlation. The clinical data is incredibly important when you're doing variant classification, looking at the patient's phenotype, we're looking at segregation, de novo events, you know, if you can have variants that, you know, are computationally pathogenic all around, but they're not causing disease. So, you know, having that information is incredibly important. And we look for, you know, there's, you know, things that may or may not necessarily be captured at this point. And some things like atypical disease presentations, you know, sometimes you have common population frequencies. I think a lot of the tools will have set, you know, standardized thresholds of, you know, 0.1 percent. This is sort of an off-the-shelf type of product. And, you know, it can miss, you know, some variants that are our founders that are more common in the population. So, again, having that clinical data is really critical to make sure that you don't miss these. I can, you know, think of a CDKN2A variant that was common in the Hispanic population that could easily have been misclassified if we're just relying on a tool. A lot of the tools are, you know, computationally derived, so, you know, that are, you know, using similar data types. And then you can have atypical variants, right? So, you know, for example, there's a variant in the FH gene that instead of causing, you know, HLRCC, which is a renal cancer phenotype, it actually causes paragangliomas and pheochromocytomas. So, you know, being aware that, you know, looking at the clinical data for that variant specific and looking for patterns does take a particular nuanced clinical experience. And unless we're, you know, the machines have enough data to learn from it, the, you know, one-offs are hard for them to predict. So there, you know, there's a lot of different exceptions. Each gene behaves differently. It keeps us very busy in our world, but I definitely do think that they are tools that can help us, that process to gather data to help us through the process. But, you know, we have to sort of take it with caution right now and work together. You know, it's a human and machine combo. Absolutely. Thanks for going through that, Tina. And, you know, speaking of collaboration, one thing that I've always been interested in is really how collaboration and partnerships across labs, academia, and clinical care sites can really help to drive our field forward. And both of you sit on multiple expert working groups and have been involved with numerous collaborations over the years. I'd love to really hear your thoughts on how this type of collaboration helps to break down barriers and drive better variant assessment approaches. Yeah, I think, you know, you hit the nail on the head. Collaboration is critical in what we do, whether it's, you know, coming up with new tools or trying to better understand, you know, genes. I think this is historically, you know, there are definitely groups that have been trying to work in this framework. You know, we've got groups like Enigma for the breast cancer BRCA genes. We've got Insight for the mismatched repair genes, who, you know, we've been working for decades on this together collaboratively. It's, you know, groups of scientists, clinicians, computational biologists that, you know, get together, share their data, work on projects, identify things that need to be, you know, better defined. And more and more of that is happening, you know, with now ClinGen, which is an NIH funded entity that, you know, we're building more and more of these working groups. They're working groups that are dedicated for defining gene disease associations. So we call them GCEPs, the gene groups. And then we have VCEPs, the variant groups, expert panels that are working on better defining the rules for their genes and diseases. So, you know, specifying the ACMG rules. And a lot of that requires those expertise. So we have people who are doing the actual functional studies in the group. We talk about, you know, where the breakdowns are, where things don't work, where things work. And part of that process is also validating that information. So it's really valuable to have laboratories involved, because we have the largest amount of data, we have the largest amount of phenotype and genotype data that can then be used to validate the functional studies, validate the in silico predictors, and then also then, you know, come up with better rules. So it's, it really is a collaborative effort. And it's a community derived sort of effort that helps with this process. And that's a process that's now FDA approved. And, you know, so it's the gold standard, and all of that information is going to be provided in the ClinGen, sorry, in ClinVar, where the variants are going to be published there, the rules are going to be published. So the whole community will benefit from a standardized process. And, you know, as I was a ex clinician, seeing patients, one of the most frustrating things, I think, is to have a sister with one result and another sister with another result, right? So creating that standardization and the data sharing, so that we have the best answers for our patients. And the more most consistent answer, I think, is, you know, the best approach in variant assessment in general. Yeah, and I mean, if I could just build on that, Tina, I think, you know, beyond participation in these ClinGen efforts, which include, you know, variant resolution efforts that are initiated through ClinGen, internally, we also do something similar, especially for rare variants, or, you know, variants where phenotype can really help. You know, we do, as clinical labs have access to much of the same information, but where data sharing is really helpful is when we have the siloed clinical information. And so, you know, when we come across a variant that could really benefit from data sharing, we take the initiative, contact our colleagues at other laboratories and ask, Hey, have you guys seen this variant? What's the phenotype of your patient? And then by working together in that way, we can come up with a resolution for the families carrying those rare variants. So, you know, some people like to say that less is more. That's definitely not true for variant assessment. More is always more. The more data, the better. And, you know, one other way that we also can contribute is by working with key opinion leaders in the field and researchers who are doing work on all of these genes that we're testing. And in that way, we can kind of get early access to information. It often takes a really long time to publish results. And, you know, some patients don't have time to wait for that to happen. So if we can get ahead of that sooner by collaborating and having these mutually beneficial relationships with these researchers, we can get there sooner. And so, you know, we really value our collaborations. And, you know, hopefully that that trickle that value really trickles down to the patients, too. That's great. Thank you both for your thoughts on that. And I love, Marcy, how you said, you know, more is more, more data always is always better. So it's great to hear how all these different groups are working together to really drive things forward. I know we're sort of running out of time here. But there is another very important topic in the field right now that I do want to talk to. And this relates to really underserved and under-tested populations. You know, we know that variant assessment can often be particularly challenging in, for example, African American, Asian or Hispanic patients. So what are some ways that we can really work together to open access to novel studies and increase data that can help us better classify variants across all populations? Yeah, very good point. You know, this has been something that we've been grappling with, you know, for quite some time. And again, back when I was in clinic, you know, we always had highest VUS rates in populations that are less tested. So you can imagine, obviously, that, you know, variant assessment is based on data. So the more data you have, the more likely you're going to classify variants. And typically due to lack of testing or, you know, lack of data for groups that are underrepresented, that, you know, this is where we have higher VUS rates. So, you know, things that we've been trying to work on, obviously, you know, the data sharing between laboratories helps, right? So, you know, we might have seen a variant a couple of times, another lab has seen it a couple of times. But when we start to pool that data, you know, what in the U.S. and also internationally, obviously, de-identify data, but putting that together, now you start to see a picture. And so that could definitely be impactful. And we've definitely seen that. And this is something that we're very strongly passionate about, about data sharing for the purposes of improvement of our variant classification. Other things that we're working on is, you know, trying to improve the agnostic type of data that's not reliant on having been seen or been seen multiple times. So what Marcy was saying previously, you know, functional studies doing these large multiplex assays that are not reliant on just, you know, doing the functional assays in variants that have been already identified, but can we do it for every position within the protein, then that data is available, even for patients who are, it's like the first time they've seen it. So then that's not going to be dependent on what, whether you've been tested or how much data there is available, you know, for that particular variant, as long as it's a high quality data, and it's been validated. Same thing with RNA. So it's DNA and protein, DNA and RNA. And, you know, again, working with more countries, you know, bringing in, improving testing, having access to counseling, and testing for broader groups, education across the communities. I think all of those things will improve our classification. Another big thing that was a huge improvement was actually these population databases. So NOMAD had a huge impact on our understanding of different, you know, variants that occur naturally in different populations. And more and more data is coming out across different ethnicities, more people are being tested, and all of that is being collected. So there's a huge initiative to get, you know, more genomes and exomes in various populations. And that has a huge impact on us in the variant assessment communities. Great, thanks, Tina, for sharing all that. It's great to hear about all the work that is being undertaken in that regard. So I think we're at time. I really just want to thank you both for participating today. I really enjoyed talking with you about these exciting topics in our field. And I just want to thank everyone for attending. We do want to keep the conversation going. So if you have any questions or want to talk to us about this more, please do visit the Ambre Genetics virtual booth and chat live with our team.

variant assessment

genetic testing

functional studies

artificial intelligence

clinical correlation

underserved populations