On The Rising Tide Podcast, scientists from the independent, nonprofit Center for Genomic Interpretation discuss with leading experts the need to further raise the bar on accuracy and quality in clinical genetics, genomics, and precision medicine. Only through improving accuracy and quality can the promise of precision medicine be realized. CGI’s ELEVATEGENETICS services are made available to health insurers and stakeholders, to help them identify the most accurate and clinically useful genetic and genomic tests, and to help them steer clear of low quality tests that have the potential to lead to patient harm and wasted expenditure.
GARLAPOW: Genetics and genomics are constantly expanding at every level: gene discovery, variant and disease associations, diagnostic technologies, and patient care pathways. At each level, we need to create checks and balances to allow for growth, while also maintaining quality, and ideally, have consensus among stakeholders in the industry. Who is responsible for these goals, and how do we navigate the exponential growth in genetics and genomics? Today, we talk with Dr. Madhuri Hegde, a co-author on the 2015 “Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology.” We start out by defining what variants are, and what variant classification means.
HEGDE: Any change in the DNA is commonly called as a mutation. So, we do not in, clinical genetics, have been using the word “mutation” but that has changed with time. So a gene in which a change is happening, we call that a variant. And that’s an effect on the gene, whether it could be a loss of function variant or the gain of function variant. And the interpretation of that variant for causality of disease is variant classification.
GARLAPOW: Next, we look at who the stakeholders are.
HEGDE: I look at the stakeholders like an ecosystem, which is interdependent on each other. Starting with the patients, the physician, the link between the physicians and the labs very often is a genetic counselor, and the laboratories itself. Then going down another layer, are the companies which are developing software for variant classification or rather automated variant classification, or even the kits and the assays that are getting approved through FDA for commonly known pathogenic variants. And then the third level being the research community which kind of sits all around this groups of stakeholders because they are doing the fundamental research identifying new genes, understanding the changes of those variants on the gene expression, on ability to cause disease themselves.
GARLAPOW: Dr. Hegde gives us some insight into how the American College of Medical Genetics and Genomics, or ACMG, created the 2015 classification guidelines as the scientific community learned more about the evolving world of genetics and genomics.
HEGDE: So all this culminated into this short time, and I call it a short time from 2010 to 2015, where this need for new guidelines was very much needed. An era that we went from 200 genes to about 6,000 genes, which today we say that there is some way that they can cause certain different clinical conditions. And with all this happening, the difficult part of it is how do you make sense of those 200 genes well-characterized, well-assessed, still having difficulty every now and then in interpretation, and then this whole new list of genes in the research community with very few publications, at times a single publication on a single gene, and how to bring that all together and interpret it. Well, it actually took us two years, two full years to get the guidelines together. And there was a reason for that, because we were doing this, writing these new guidelines in the middle of implementation of new technology such as next generation sequencing, and going from panels to exomes and genomes, and the debate of what we are going to do in a clinical laboratory. Now another thing which is very important to remember is that the guidelines were meant as a general guideline, a very high level, they are the umbrella guidelines to use when going and drilling down into different genes. The American College of Medical Genetics and Genomics has a global impact, and people look up to ACMG to give those guidelines out. I think, just overall, if you had to grade the guidelines, that they have been a huge success.
GARLAPOW: We also discussed some of the hopes Dr. Hegde has for the next guideline update from ACMG.
HEGDE: I think achieving consensus is not an easy thing. I think that is probably the most important thing we have learned from the 2015 guidelines. Especially in this complex world of genetics and the new genes that are getting identified and reported. I think the next version hopefully is going to cover several things. The first thing being that there is just so many guidelines now, gene-specific guidelines, and we should still be able to bring these guidelines together in such a way that the umbrella guidelines can be modified to be used as a more generic guideline so that labs don’t have to rely so heavily on using every single different gene guideline. Then you are starting to rely on software which are getting sort of in the market to take those single gene guidelines and accurately interpret them. The question is whether that can be really done and we should rely entirely on the software or not.
GARLAPOW: Artificial intelligence’s role in variant classification has notable, critical limitations.
HEGDE: The last two years have been interesting, right, with AI, that there’s so much hype about AI. Definitely AI is going to be very useful going forward, so it’s not like we’re all saying that, “Oh this is very, very dangerous and don’t do it,” but the human aspect of that cannot be removed because variant classification is a science but also an art, and there is that human piece of it which you just cannot remove. Variant classification does require human intervention, and it is not something you can rely on just looking at databases.
GARLAPOW: Given the ever-changing landscape of genetics and genomics, Dr. Hegde challenges clinicians to elevate their understanding of genetic testing and interpretation processes.
HEGDE: The question the genetic counselor should be asking the laboratory is who is doing your interpretation, how do you do interpretation, what data do you have access to to doing interpretation? When you get a clinical report, talk to the laboratory, ask questions. What testing was done, what are the missing holes, are there proper disclaimers or not? Absolutely critical when taking a clinical report and applying it to patient care. And it’s very important to read the details at that point because you are actually going to affect someone’s life. I think in the whole scheme of things of genetics and genomics we forget one fact, that we are still very much in the early days of how genetics is going to get applied to the population in general.
GARLAPOW: Dr. Hegde’s vision for the future of clinical genetics and genomics is optimally propelled by collaboration.
HEGDE: To raise the tide, I think we are heading in that direction definitely, but what we need to do is collaboration. Today, no one can work in a silo. Collaboration, collaboration, collaboration, and education. Like this podcast, is absolutely critical and working with all agencies across the board.
GARLAPOW: I’m Dr. Megan Garlapow with the Center for Genomic Interpretation, and you’re listening to The Rising Tide podcast, where we learn from experts about improving the accuracy and quality of precision medicine and clinical genetics and genomics. Please note that this podcast does not provide medical advice and is not intended to be a substitute for professional medical advice, diagnosis, or treatment. Always seek the advice of your physician or other qualified healthcare provider with any questions you may have regarding a medical condition. Additionally, comments of The Rising Tide’s guests are their own and do not necessarily reflect the position of the Center for Genomic Interpretation.
Today I am joined by Dr. Madhuri Hegde, VP and Chief Scientific Officer of Global Laboratory Services at Perkin-Elmer, which she joined in October 2016. The focus of her clinical work is the development and implementation of high throughput next generation sequencing (NGS) strategies for genomic disorders using sequence capture technologies (robotics, clinical exome and genome sequencing, and oligonucleotide array platforms). Dr. Hegde has extensive knowledge of CLIA regulations related to laboratory-developed tests or LDTS and NGS and a strong understanding of the current reimbursement landscape and new CPT coding guidelines. Previously, Dr. Hegde was with Emory Genetics Laboratory in Atlanta, Georgia where she was Professor Human Genetics and Pediatrics, and Executive Director and Chief Scientific Officer of Emory Genetics Laboratory. Madhuri received a Bachelor of Science and a Master’s of Science from the University of Bombay, India and a PhD from the University of Auckland, New Zealand. She completed postdoctoral studies at Baylor College of Medicine, and is American Board of Medical Genetics certified in clinical molecular genetics. She was a member of the committee that established the 2015 American College of Medical Genetics (ACMG)/Association for Molecular Pathology (AMP) variant classification guidelines whose implementation we will be discussing today along with other important areas of clinical genetics and genomics. Please note that Dr. Hegde’s views are her own and do not necessarily represent the views of her employer. Additionally, some of what we discuss is speculative in nature and/or describes use of approaches that have not necessarily been validated or approved.
In this episode of The Rising Tide podcast, we discuss, among other topics, one of the steps in the process of clinical genetics and genomics, called variant classification. So when a patient’s clinician orders genetic and/or genomic testing to assess for disease-causing or pathogenic differences or variants in the inherited germline genetics or the changed somatic genomics of a patient’s DNA, clinical laboratories get a list of variants from the sequencing or panels or however they’re assessing this. Then, a variant classification scientist determines whether a variant is pathogenic or likely pathogenic, which is a positive result, a variant of unknown significance, called a VUS, or a likely benign or benign variant, which is a negative result. The variant classification scientist then creates a report that they send to the clinician. This process of classifying variants can be complex, intense, and difficult. Some have described it as the “Achilles heel” of clinical genetics and genomics. Many opportunities exist across the entire process for error. Because of this, the ACMG and AMP created the guidelines for variant classification with the most recent guidelines from 2015. As I mentioned, Madhuri was part of the committee that created the 2015 guidelines. The guidelines are currently being updated and are expected to roll out later this year. I have had the pleasure of working with Madhuri before and had conversations about variant classification and other areas that were very specific and technical and a whole lot of fun. Here we hope to open up the discussion to a broader audience. Of course, if you have any questions, please feel free to reach out to me or any of the team members at the Center for Genomic Interpretation. Madhuri, thank you for joining me today.
HEGDE: Hi Megan, and thank you for having me on this podcast.
GARLAPOW: Of course. So in your terms, if you were speaking with a friend who is not a variant classification scientist and has no experience there, what is variant classification?
HEGDE: I think that’s a great question to start with. The statement of “variant classification” is so widely confused, so let’s just start with basics which I tell myself and divide it into three categories, the first one being that any change in the DNA is commonly called as a mutation. So, we do not in clinical genetics have been using the word “mutation” but that has changed with time. So a gene in which a change is happening, we call that a variant. And that’s an effect on the gene, whether it could be a loss of function variant or a gain of function variant. And the interpretation of that variant for causality of disease is variant classification.
GARLAPOW: Okay. I think that’s very clear, and I used to also use the term “mutation” quite a bit. Perhaps because of the advent of X-Men we moved away from that language. “Variant” also sounds nicer. “Mutation,” even if it’s benign or likely benign, I think the word “mutation” has that connotation of something much more dangerous.
HEGDE: Right or something bad, and you know we all tend to use that interchangeably, but more and more and I have done that myself too, I keep telling myself that we have to use the word “pathogenic variant” or a “benign variant” which is a causality.
GARLAPOW: Yes, yes definitely. So why does variant classification even matter? Who should be interested in the quality and accuracy of variant classification? Why should they be interested? Who are the stakeholders here?
HEGDE: I look at the stakeholders like an ecosystem, which is interdependent on each other. Starting with the patients, the physician, the link between the physicians and the labs very often is a genetic counselor, and the laboratories itself. Then going down another layer, are the companies which are developing software for variant classification, or rather automated variant classification, or even the kits and the assays that are getting approved through FDA for commonly known pathogenic variants. And then the third level being the research community which kind of sits all around this groups of stakeholders, because they are doing the fundamental research identifying new genes, understanding the changes of those variants on the gene expression, on ability to cause disease itself. So it’s a group of several stakeholders, but they can be independent and also interconnected.
GARLAPOW: Okay, that’s very interesting and very accurate. So understatement of the year: genetics and genomics is evolving rapidly. The whole landscape is changing just so quickly. Clinical genetics and genomics drive significant decisions on treatment for patients and healthcare spent. How did these factors affect development of the 2015 guidelines?
HEGDE: So in the last 20 years or so, as you just said, you know the technology is rapidly evolving around us. I still remember in my postdoctoral days and ACMG training days, we were still doing Sanger sequencing and at that time, the guidelines had just started coming out. There was a 2002, a 2005, ‘07 guideline which were really succinct in the sense of definition of how to interpret a variant, and then there was this big gap in between where a committee was put together to write the new guidelines largely because of about 2009 and 10 when next generation sequencing started getting used in clinical labs. It was very clear at that time that we really needed a new guideline. Also, as we all know, we suddenly went from identifying about 10 genes per year to about 100 or 200 genes per year, and that escalation really needed new guidelines. So the idea behind the 2015 guidelines really was to bring together physicians, laboratory geneticists, genetic counselors together to write this new guideline which will reflect on the new use of new technology and how to interpret those variants in this new era of next generation sequencing and where we were going with genetics and then looking at genomics as a new sort of discipline as well.
GARLAPOW: So, you did describe this but can you do so again. The landscape of variant classification prior to the release of those 2015 guidelines, it sounds like it was also evolving really rapidly and there was sort of that inflection point of its evolution around 2009-2010, but can you describe it in a bit more detail?
HEGDE: Sure. So the period of 2010 to 2015 becomes very important here that what actually happened and the need for these new guidelines. And I remember when this committee was put together, the expectation was that, you know, maybe in three or four months or six months after starting this committee we will be able to put out the guidelines, but it actually took us two years, two full years to get these guidelines together. And there was a reason for that, because we were doing this, writing these new guidelines in the middle of implementation of new technology such as next generation sequencing, and going from panels to exomes and genomes, and the debate of what we are going to do in a clinical laboratory. Also, one important thing to remember here is next generation sequencing is not the only technology used in clinical labs and it never will be probably. We still have Southern blots and we still have the need for traditional methodologies. Alongside of that is also, as the industry started getting more and more interested in genetics, we had many startup companies. You have companies developing extensive software for looking at different variants and then, as I mentioned earlier, the assays or the kits where certain defined mutations would– I used the word “mutations”– pathogenic variants would go into that particular assay. I think a good example of that is ACMG list of the cystic fibrosis pathogenic variants, where that list kept changing a little bit of one or two variants here and there depending on their effect. So all this culminated into this short time of, and I call it a short time, of 2010 to 2015 where this need for new guidelines was very much needed. And not that the old guidelines did not address a lot of this, they did address. In fact, the old guidelines, it’s some of those definitions and sort of the layout of how to go about interpreting a variant are still widely used, but bringing that all together under one scheme if I can say that and also using that scheme and applying it with different definitions was very critical at that time. I think the committee really did a good job in achieving that goal.
GARLAPOW: Wonderful answer, thank you Madhuri. So you did just touch on this when you described the landscape prior to the 2015 guidelines, but it’s worth commenting on again. What was intended with the 2015 guidelines? Why were they needed and what did the committee hope to accomplish? Why was this so important, and I’m asking this again because it’s such a critical question, understanding why these guidelines had to happen? It’s so important here.
HEGDE: So see the other thing, and I think probably sort of looking at it from the medical side of the interpretation of the variant is, you know, we are talking of 22,000 genes in the human genome. And when this evolution started with technology, we were at about 200 genes or so and which we knew really well, were assessed really well, and even with that we knew that it was difficult to come to a conclusion about a variant. Now suddenly, we are looking at this era that we went from 200 genes to about 6,000 genes, which today we say that there is some way to infer their causality or that they can cause a certain different clinical condition. And with all this happening, the difficult part of it is how do you make sense of those 200 genes well-characterized, well-assessed, still having difficulty every now and then in interpretation and then this whole new list of genes in the research community with very few publications, at times a single publication on a single gene, and how to bring that all together and interpret it. Many labs have this difficulty because when you are finding a change which matches the phenotype or the clinical presentation and you’re trying to sort of do the variant classification, it gets very difficult when not much is known about the gene. Now another thing which is very important to remember is that the guidelines were meant as a general guideline. A very high level, they are the umbrella guidelines to use when going and drilling down into different genes. So from the start of variant classification era, we can say about 2002-2003 where it really actively became very important to have guidelines, they were always meant and even today, the 2015 guidelines were meant as umbrella guidelines to use as a guide to developing strategies to approach different genes causing a variety of different conditions.
GARLAPOW: So you have umbrella guidelines for this increasingly complex space that’s covering really complex diseases and disorders, and it’s supposed to support and help. So, how did implementation of these 2015 guidelines proceed? What was expected and what obstacles and challenges arose?
HEGDE: It was, as I mentioned, the committee was absolutely a fantastic committee to work with. We debated a lot, we had several working sessions where we would go back and forth with ideas on how we can present the classification scheme in a more succinct manner, but also looking at it from the other side that the laboratory is going to take this, the expectation being that the labs are going to use these guidelines to interpret a variant and then you know the next level being that clinical report going out to a genetic counselor or a physician who is going to take that report and actually talk about that report to the patient. So these are all different ways of looking at it, and how you can make it easier for this entire chain to understand a variant and its effect or is it really causative of the patient’s condition. So the first part of it is obviously the publication of those guidelines, and as I mentioned it took us two years to come to that manuscript and publish it largely because of as I just mentioned the changing landscape of technology and our ability to understand genetic diseases. And the demand was so high, there was a significant amount of ask for it, so we were glad that we were able to get it out. But after the guidelines were released, the implementation of it and how many clinical labs would actually sort of absorb them and put them to use was also very important to understand. At that time, when the guidelines were released, I was also on the College of American Pathologists (the CAP) biochemical and molecular genetics committee, where I used to assess the proficiency testing results for sequencing. My primary responsibility was to look at the sequencing proficiency testing results and it was really a very sort of nice to see that the labs picked up, the labs were so hungry for these kind of guidelines, that they picked up those guidelines very quickly and started using them. We are talking of guidelines which are not limited to just use in U.S., right? American College of Medical Genetics and Genomics has a global impact, and people look up to ACMG to give those guidelines out, so the impact was really wide in almost, across the globe. I do a lot of CAP inspections as well, and it was really good to see, doing a local or national and international inspections that labs were starting to use these guidelines. Of course a lot of questions started coming. So the most immediate thing after the publication was the critique around maybe they could have done this and done that, and some critique around especially BRCA1 and 2 at that time was so important that we had not specifically dealt with gene-level interpretation of variants. So some clarifications followed after this publication was done, where we sort of, and as I said they were meant to be umbrella guidelines. And if you see most of the committee members who give presentations after the publication, they really talk about the guidelines being the umbrella guidelines and then developing gene-level guidelines. And the good thing that happened right after that was the award of the ClinGen grant where all this started being implemented at a much wider scale and sort of being drawn and absorbed globally.
GARLAPOW: That’s really, really interesting. And just a reminder for our listeners, BRCA1 and BRCA2 are two genes that can harbor both inherited germline variants or develop somatic variants that can, when those variants are pathogenic, they can increase the risk of developing certain cancers. When those are inherited variants, it’s a hereditary cancer. And we discussed this in BRCA1 and BRCA2 to some extent on the first two episodes of The Rising Tide podcast, it’s good to circle it back in here with Madhuri. So what successes did you observe and/or experience from implementation of these guidelines and what surprises were there? You talked about how labs almost immediately started picking these up, they seemed like super hungry for them, so to speak, so what successes and surprises, pleasant surprises, were there along as well?
HEGDE: I think just overall, if you had to grade the guidelines, that they have been a huge success. I think it’s kind of caused that awakening in the community of genetics and even at the people who are involved only in just doing basic research. Now I’d look at these guidelines. So I consider that a huge success, because even though writing the guidelines was really meant for clinical laboratories to use those guidelines for interpretation of variants, the implementation and how they were absorbed was globally and across all different levels, and by that I mean again where I started is the clinical labs you have the vendors who are developing this software or assays, and then also the research community who also started looking at these guidelines. So overall it’s been a huge success, and the implementation and absorption of it globally. Even small laboratories. I’ve done inspections in China and Taiwan who were also looking at this. It was really nice to see that going for these inspections and them using these guidelines. So that is a great part of it. Some of the surprises, and I will stick to pleasant surprises–
GARLAPOW: We can talk about the unpleasant ones too! We’ve got room for that, but let’s start with the pleasant ones.
HEGDE: Some of the pleasant surprises really were, you know, we had a lot of databases for gene-specific databases before the extensive guideline came out, and I started seeing the gene-specific databases also implementing the terminology. And as I mentioned as we started the podcast, where the terminology becomes so important going away from “mutation” to “variant” to “pathogenic variant” and breaking it down from there, that was really good because I think that just tells you that the community has started coming together to create this consensus at least on the terminology side of it. So that was really good to see. The other sort of pleasant surprise for me personally, because I am a laboratory director, I do clinical interpretation, is labs coming together to form groups to do variant reinterpretation. And that became just such an active exercise especially in the period of 2015 to 2018, and you will see a lot of publications where variant reinterpretation and laboratories competing with each other for business, they’re collaborating with each other to do variant reinterpretation. I think that was just a fabulous experience.
GARLAPOW: That’s really beautiful too. We actually have an upcoming episode of the podcast on variant reinterpretation and reclassification as well. So, what were some of the unpleasant surprises?
HEGDE: I think the unpleasant surprises really fall into like two or three groups, if I can say that. The first one is the immediate reaction, right, that they expected so much more from the guidelines and the critique that this was not covered or that was not covered in the guidelines. For example, as I mentioned, the confusion at the start that, “Oh these are not gene level guidelines and we’re applying it to my favorite gene and they don’t apply directly to it.” So that was kind of the most immediate reaction to the guidelines itself. The second piece of it in terms of the unpleasant reaction to the guidelines was the sort of the putting that algorithm together in a more easy way for labs to use. So you will see a lot of publications that came out in that period also where people started developing computer-based algorithms to automatically call a variant. There was so much discussion on that because it is not just a science or algorithm you can put together, you know, variant classification is also an art. There is a place for individuals to think through the variants and actually do that accurate correlation with the clinical presentation. That took a lot of beating, if I can say that, because it was just so many publications that came out. And even today, they are still coming, and it’s not accurate. And I think it’s important to understand that we were not trying to take away the entire sort of the art side of variant classification and eliminate all lab directors if I can just say that and make it entirely driven by computer, by software. So these were kind of the early, unpleasant experiences of trying to explain this, that what we had put together was to be used as a generic guideline.
GARLAPOW: Yes, and actually we’re going to discuss that in further detail in a bit, but I’m really glad you brought it up here so that we can re-emphasize it. Artificial intelligence is being increasingly deployed without thought toward the art side of variant classification and it is so important to consider the implications if you remove the variant classification scientist and that training and expertise and the art side that is needed to this, so I’m so glad that you brought that up already. So we are six years from the 2015 guidelines and I don’t know about you but this is the most rapidly evolving field I have ever worked in. And so the guidelines are being updated. That is awesome. What are your top recommendations for the updates? Why?
HEGDE: Again, another great question. So 2015 till now, I think that there’s so much that has happened, and as you just mentioned I think we are very fortunate that we are working in this field and just the rapid development of technology and trying to bring it to clinical practice, it’s just been an outstanding experience. The new guidelines, and I think before we talk about the new guidelines, we have to talk about this in-between period of what has happened. The ClinGen grant was awarded by NIH. Several groups have worked, really extensively, volunteered their time for variant classification, databases have evolved, a lot of sequencing has happened, and so much of sort of variant accumulation if I can say that, from different sequencing, large sequencing efforts, have contributed to the database. That is one sort of massive development since the guidelines where, because at that time we were using just the dbSNP or the HGMD type databases. And then, in this period also you had the different ClinGen groups starting to publish. So I co-chaired the PTEN committee and we have a PTEN gene-specific guideline. So now we are really going down to a single gene or a group of genes and those guidelines are starting to come out. And what have these guidelines really done? It’s taken the gene, the correlation to the disease itself, and modified the guidelines based on the incidence of the disease or the penetrance of the variants, the clinical spectrum of the disease itself, so the PTEN guidelines have modifications and the percentages or the incidence of the disease and then the allele frequency. And different groups have done it differently. There’s an epilepsy group, there is a neuromuscular group, there is a cardiology group, so all that is great, and I think the next version hopefully is going to cover several things. The first thing being that there are just so many guidelines now, gene-specific guidelines. And realistically thinking, as a clinical molecular geneticist, I do single genes, panels, exomes, genomes on a daily basis. The need for speed versus the need for accuracy and what is important. Obviously, the need for accuracy of the clinical reports, and for that reason, it’s really important that we are still able to we should still be able to bring these guidelines together in such a way that the umbrella guidelines can be modified to be used as a more generic guideline so that labs don’t have to rely so heavily on using every single different gene guideline. I mean, if you can imagine the world of even 6,000 genes, it’s almost impossible to do that. Then, you’re starting to rely on software which are getting in the market to take those single gene guidelines and accurately interpret them. The question is whether that can be really done and we should rely entirely on the software or not. So that’s one part of it. The second part of it is actually looking at the variant itself. And I just recently published a paper with Joe Shen and Clara van Karnebeek in “Genetics in Medicine” is this reverse interpretation of a variant and how we can bring it in clinical practice, and by that I mean, most of the reports today have variants of uncertain significance. And in a recessive condition, if you have reported a pathogenic variant and a variant of uncertain significance, and finally the clinical lab is not the lab which sees the patient, it is the physician who is taking that clinical report and interpreting it for clinical care. If they have started some treatment or management, a plan based on that clinical report, because the gene fits the clinical presentation, and the patient has actually responded to treatment. I think that’s your in vivo proof that the variant of uncertain significance at that time could be reclassified. We rely heavily on in vitro experimentation of the functional genomics and prediction software, such as the computer prediction on pathogenicity versus now you have in vivo. Now to a certain extent, the umbrella guidelines do cover it, but it does not assign it a separate category. So I think these, kind of, new thought processes are very important to be included in the new guidelines.
GARLAPOW: Yeah I think that is so important to know if this real human patient responded to the treatment that was guided by that clinical genetic/genomic report that the clinician received. And also I think it’s really important, we call these diagnostic clinical genetics and genomics, but to remember that it’s the physician, the treating physician, who makes the diagnosis. So following onto that, what need to be the top priorities for adapting the updated guidelines and why? What did we learn from the implementation of the 2015 guidelines that could help us now as we’re approaching the roll out of the updated guidelines?
HEGDE: So I think achieving consensus is not an easy thing. I think that is probably the most important thing we have learned from the 2015 guidelines, especially in this complex world of genetics and the new genes that are getting identified and reported. It’s very, very difficult not just to achieve consensus, but to educate at all these different levels, so the stakeholders, and making sure that they’re applied appropriately at each individual layer and by that I mean the laboratory doing the testing, the geneticist doing the interpretation variant, to the genetic counselor, to the physician, and what the patient takes home understanding of their disease and what to do next for their own care or their family’s care. So the learning curve has been huge in terms of absorbing these reactions and how to move forward with this. So I look at two things as being very important. One, is the entire ecosystem coming together and not really having a common or rather we are never going to reach consensus. Let’s just agree to that, because we all come from different backgrounds, operate in different settings, operate in different cultures. You know, you cannot expect the culture or the approach to genetics in a country like India to be the same as in the U.S.. Understanding that and taking that approach forward is education, trying to at least come to some consensus, but most importantly a common approach of how we are using these guidelines is going to be critical.
GARLAPOW: Okay. Thank you. So I’m going to define ClinVar by reading its definition from NCBI because I’m about to ask a few questions about ClinVar, but before I ask the questions after I read this, I’d love to hear ClinVar defined in your terms as well. So according to NCBI, “ClinVar is a freely accessible public archive of reports of the relationships among human variations and phenotypes with supporting evidence. ClinVar thus facilitates access to and communication about the relationships asserted between human variation and observed health status and the history of that interpretation. ClinVar processes submissions reporting variants found in patient samples, assertions made regarding their clinical significance, information about the submitter, and other supporting data,” end quote. So it’s a resource and database of sorts for variant classification. Is there anything else do you think I should add for ClinVar that isn’t covered by what I just read from NCBI?
HEGDE: You know, going back to the start of the podcast, where I sort of defined that variant classification, that layer between a variant and its effect on the gene and the causality, I think that subtle but very important difference is critical for adding to the definition of ClinVar, because the effect of the variant on the gene if it is loss of function, let’s assume for a second, is not necessarily going to translate into pathogenicity or it being disease-causing. There is a whole body of data in between these two things which we call the functional aspect of it that needs to be understood and incorporated.
GARLAPOW: Absolutely, absolutely. So ClinVar is a resource and database of sorts for variant classification, where if you have a variant that’s described as a VUS you can see the evidence that is up there in some variant classification categories. For some variants, you can see discordance between different groups on its interpretation. What have been ClinVar’s greatest outcomes and successes in this world of variant classification and what areas need improvement?
HEGDE: So ClinVar did exist to a much smaller footprint before the ClinGen grant was awarded, and I think one of the big successes of ClinGen and therefore ClinVar encouraged so many webinars and meet-and-greets with labs early on after the 2015 publication to encourage labs to submit their variants to ClinVar, and that was a massive exercise. Remember, all labs are also competing with each other, so to go from that. For academic laboratories, it was easy to come to that conclusion that, “Yes, we will submit all our variants,” but when we are talking of private commercial labs, giving away that intellectual know-how or knowledge about a variant was not easy to convince. And today you will see in ClinVar almost across the globe, whether it is academic labs or commercial laboratories, they are now submitting to ClinVar. I think that’s a huge success to convince the community to come together, the testing community. The second thing is, ClinVar has gone through significant modifications. Early on the feedback was, “It’s very difficult to upload a variant to ClinVar,” versus now they have made that process much easier that it’s now easier to get those variants from any lab uploaded. You can set up some quarterly or half yearly plans to submit variants to ClinVar. And I think the third thing is NCBI folks, along with the ClinGen groups, have listened to what the consumer, which is a laboratory or a physician, are wanting to see in ClinVar, is the display of a variant has changed significantly. If you look back about five years ago how the variant submission looked, the display and now what is there, they are starting to collect evidence, adding multiple layers of evidence. The one thing which really needs to be done, so this is the successes, but is going to your latter part of your question is what needs to be done. The issue is, as you just mentioned, that the discrepancies between the labs are still there. So a lab which has submitted in 2016, let’s say, and today, if a lab goes in, at that lab could I call it a variant of uncertain significance and the latest version could be pathogenic or benign, either way, what happens is that you really need to take a look at those variants and bring it down to this most recent classification because for a laboratory with many directors doing interpretation, it’s very easy to understand that the variant classification also evolves over time. Versus, for newly certified directors or small laboratories, it’s not easy to understand what this all means, which one to follow. So I actively tell my fellows that look back at how many times that variant has been classified over a number of years and what evidence is available. But the display comes out as variant of uncertain significance and if you sort of click on it, it gives you that full range of variant classification that I think that should be the top priority to clean up this extensive accumulation over 10 years for certain variants. So that’s number one. The second thing is I think in terms of improvement, including that in vivo data is going to be very important. We heavily rely on functional genomics and the in vitro data, which is going to remain and this is going to be the primary way of interpreting or having functional evidence, but adding that is going to be important. And I think the step after that is this diversification into, they still have it today, is the risk alleles and the polygenic risk scores because labs are starting to report risk alleles and polygenic risk scores, so that’s going to be hopefully the next layer that gets brought into ClinVar as well.
GARLAPOW: Wonderful. Thank you, Madhuri. So recently, the FDA has established FDA-approved variant classifications in ClinVar. What this means is that some variants have FDA approval based on committee concordance, so a committee of experts. What are your thoughts on this? What does this and then after your thoughts on this, what does this mean for very rare variants and what about for polygenic risk scores which you just mentioned?
HEGDE: Right. I think it’s a great thing that FDA has noticed ClinVar, and this is a part of the success story as well that we are starting to get FDA into this fold of getting approvals in place for this variety of things we are doing in clinical genetics, but the effect of that approval is confusing for people. I’ve been to many meetings where someone will suddenly say to me that, “Oh it’s a FDA-approved database.” No, it is not a FDA-approved database. It’s a variant level approval where ClinVar and ClinGen have put together the star level system of how much evidence you have for that variant and how many stars it has, so I think that’s the first level of clarification which is absolutely needed because a FDA-approved variant, it is not a FDA-approved database, and how do we explain that to the testing community is going to be critical. Now what does it mean by being a FDA-approved variant for pathogenicity or being benign? It’s important because the companies which are looking at these variants and trying to bring smaller panels together or developing reference sets for labs to use for setting up assays. This level of classification is very important. So it’s a good thing, but I think it needs to be a little bit more defined in that sense. The risk scores and risk alleles and the polygenic risk scores and defining that within the system is not a trivial thing to do, and whether we should report that in the clinical setting today itself is an ongoing debate because explaining that to someone at the risk or your odds ratio of getting a disorder or your risk of getting a certain condition is not a very easy thing to do because it’s a very statistical analysis of the data itself, but there are guidelines coming out for that as well. Labs are seriously starting to look at it. And hopefully in ClinVar, ClinVar does have by the way some risk values, but I think a little bit more realigning of how that data is viewed by the user is going to be absolutely critical.
GARLAPOW: Okay. I think that’s really important too. And yes, risk scores are hard to communicate accurately, so I’m really grateful that you brought that up too. So we have two distinct areas of genetics and genomics. We have clinical, used in diagnosis and clinical treatment of patients, and then we have basic research, the research side of it. Increasingly, the lines between basic research and clinical research are becoming blurred. Why is this happening? Why does this matter and what does this mean for clinical genetics and genomics?
HEGDE: That’s again a very, very important question. In fact, I have given many presentations where I say that the lines between the clinical genetics and research are getting blurred very fast. So the first thing that happened with the new technology is that clinical laboratories started reporting new genes. Traditionally, we expect basic research labs, which are collecting a cohort of patients and going after identifying the gene doing the functional basis of that gene identification, and then a publication comes out. That has been the normal, traditional pathway. That the publication will be there, then the clinical lab will decide whether this deserves to be in a clinical setting, can you develop an assay, then starting to look at what is the best approach to develop a assay or to test for it. That is just totally now blown out of the water, if I can say that, because clinical labs are identifying new genes, because so much more testing is happening there’s so much more accumulation of data, and they are now collaborating with the research labs to report new genes. I mean many, many diagnostic testing labs have published new genes from their testing populations, so this is a very different scenario now that very rarely you will see now a publication coming from a basic research setting. So that’s one important thing that has happened, but we still heavily rely on the basic labs, research labs. Two reasons, right? The functional genomics aspect of it and this is one thing hopefully ClinVar will look at establishing, or ClinGen rather, is not just having the gene and the variant but which functional domain or which functional assay can a clinical lab rely on. This is one question that comes very often in when I go to conferences or meetings, that, was the right functional assay used or not to show that the variant effect is in whichever direction, but that is probably one of the most important things that the research in the clinical labs need to kind of come together and hopefully implement it through ClinGen into ClinVar, that this functional genomics aspect of it where the research labs are going to play such a big role. For us to implement, and I think eventually that functional genomics aspect is going to start to come into the clinical labs as well. For example, RNA sequencing, to understand the effect, clinical labs are starting to do it now. And the last, but not the least, is the basic research labs are going to absorb this data generated by clinical labs in and submitted to databases to do the basic research aspect of it, and there’s just so much more to be done. I think we have plenty of genes yet to be identified. Publishing negative data is also important. I think this is one thing which kind of just gets lost in all these discussions that, “Oh, we want to identify a new gene.” How often do we look at negative data, is where a gene was published to be causative of a particular disorder or a variant reported and you see that in OMIM there are many such legacy variants. How often do you see publications which show or say that, “Oh, this gene no longer is associated.” We had done a publication when I was at Emory for the MYH8 gene to show that it took us a long time. This is probably one of the longest times it has taken for any of my publications to get accepted, because no one is interested in negative data. Everyone wants this whole positive data, so I think these are the things to consider as going forward.
GARLAPOW: Absolutely. When I was in academia, I would joke with some of my colleagues that what we needed was the “Journal of No.”
HEGDE: No, I mean I think this is very important data and this needs to get published.
GARLAPOW: Absolutely, absolutely. So what knowledge gaps exist in variant classification? How do these gaps contribute to inaccurate variant classification, and how does this affect false positives, and what can be done?
HEGDE: So gaps in variant classification. So, I think this flows nicely from the previous question you asked me is this whole big piece of functional analysis of a variant. Now, we are doing a lot more genomes. I mean, my lab has slowly started seeing the trend. So I’ve been at Perkin-Elmer for about four years now, and the trend is shifting from exomes to genomes very quickly. In fact, in the last six months we have been doing a lot of genomes now. So the first gap, probably, which sort of is in our face now, is because we are doing genome level clinical testing, there is this intergenic and intragenic sequence that we need to pay attention to. How are we going to interpret that? What needs to be done to look at those variants? And hopefully the new guidelines will have some comments on that as well, of what to do when you’re doing genome sequencing. Why do people actually do genome sequencing? The reason for that is the unbiased sequencing aspect, but your ability to look at all the variants in the gene. So this is probably one of the most important things in variant classification gaps that need to be addressed. The second is, again I think I already mentioned this, is the functional assays used for variant classification. And you will find that there are many publications where a particular variant is taken and you can see the difference in the interpretation or based on the assay they have used to classify the causality of that variant. That gets very difficult to interpret for the clinical lab where you’re doing, you know, any typical director is reporting 10, 20, 30 cases per day to actually drill down to that level whether the right assay was used or not, or what is the right assay. So I would say more than focusing on gene level classification systems and schemes, I would rather have some more focus on which is the right functional assay to use, and is it right or not, so that the lab directors can assess it in that particular aspect. And, you know, just thinking of it going forward, the gaps in variant classification. I think these are gaps which we can address, like going to genome or the functional genomics part of it, but what do you do when you report the variant? And I think that comes down, unfortunately, little bit to competition between labs. Is a lab over-interpreting a variant, or what is their approach to variant classification? You will notice that many labs today have published their own scheme of variant classification on their websites. I often get asked this question, what is your rate of variant of uncertain significance? And there is no such thing. There is no such thing. The question the genetic counselor should be asking the laboratory is who is doing your interpretation, how you do interpretation, what data do you have access to to doing interpretation? So let’s say I’m doing a genome case from the Middle East. Do I have access to their variant database? What variants are common in that population and does my lab do the appropriate things for variant classification, and am I open to discussion for that? I think that is what is creating this problem of false positives. False positives at the technical level is one thing and then the false positive at the interpretation level. Right now we are talking about false positives at the interpretation level, so I’ll just stick to that, and the over-interpretation can lead to a lot of problems. Clinical genetics and labs have changed so much, right? When I was at Baylor, it was just an evolving field. There were no large sales teams and marketing, and all that just did not exist in 2003-2004. Maybe one or two small commercial labs were there, but now the field has changed so much where, very often your representative going to meet a certain position is going to get asked what is your positivity rate? That is not what should be asked. I think the qualification applied to a laboratory is how do they do the work that they are supposed to do?
GARLAPOW: Yes. Is there stringency? Is it done rigorously? Are they integrating the science and the art? I have heard of a lab that biases toward likely benign instead of VUS to decrease their so-called VUS rate, because they’ve become aware that clinicians often feel uncomfortable seeing a VUS. And I cannot emphasize enough, if this is actually happening, what a disservice this is to patients in the end, because they then get a report of say a likely benign variant and if that variant becomes reclassified in the future from VUS, where it should have been, to likely pathogenic, that is just such a missed opportunity to deliver on excellent, accurate care which is what patients deserve.
HEGDE: Yeah, and you look at the other end of that spectrum too, that overcalling a variant. I mean, I always tell my fellows that when they come and they’re debating, call it likely pathogenic or pathogenic if you will test your own sample, and you will be able to confidently test your sample for that gene, and if that variant comes up and call it likely pathogenic. Because the next thing that happens when you call a variant pathogenic is it triggers a prenatal very often. And, if you don’t have that confidence, leave it in the area of variant of uncertain significance. I think in the whole scheme of things of genetics and genomics, we forget one fact, that we are still very much in the early days of how genetics is going to get applied to the population in general. We are in this infancy right now, and let’s not jump into this rigorous sort of definition of classification when we don’t know something we don’t know, we’ll go find them and then we’ll reclassify it. And that’s why that reclassification and reinterpretation becomes important.
GARLAPOW: This moves so naturally into one of the questions that’s a bit further up in the discussion guide I shared with you, so I’m just going to go for it. Let’s set up an example here. Let’s say you are some sort of a pediatric specialist and you have this three-year-old boy and he is so sick, and he’s so sick in a way that phenotypically matches with a rare genetic disorder. You see that for sequencing, and it comes back, the report comes back with a VUS variant in that gene that is expected to be causal when it’s a pathogenic variant of this rare disorder. And there’s a treatment option for this patient, for this little boy, and it could help him, and you’re the doctor and you get this VUS you don’t know, but you have clinical suspicion that it is this rare disorder. What do you do? What role does compassion play here? How can we address compassion in the world of variant classification?
HEGDE: That’s a great, great question. So last week, I gave a webinar in a pediatric neurogenomic symposium and the title of the webinar was, “How Should a Physician Look at a Clinical Genetics Testing Report and What to Do Next.” How to read the report, first thing, right? Because before we talk about compassion I think the most important thing is, how should the physician read the report and interpret the report for patient care? And it’s very important to read the details at that point, because you are actually going to affect someone’s life at the end of it. Looking at what the lab did, looking at the coverage of the assay itself, and that’s where genetic counselors play such a big role in this, that they have that understanding to help the physician in a busy clinical practice to look at whether the appropriate testing was done or not. Coverage matters. Whether what is a gene characteristic matters. What is in the variance spectrum of the gene matters, such as Duchenne muscular dystrophy is a good example where deletions, duplications are the major events. Little things like that are very important. So that’s the first step. And then now looking at the compassion side of it, it really depends on the disorder, because finally the physician has to make some decision about what care can be given. And there are a variety of ways of looking at it. Obviously, counseling plays a huge role in making the family understand what this disorder is, what this variant means, but especially when there is a variant of unknown significance it becomes even more difficult. Now metabolic disorders, we will just take one category which is metabolic disorders. The approach is different here because if something is not going to cause any harm by modifying a diet, let’s say, or giving some supplements that the patient might actually improve, I think we have seen this in many examples. I think the newborn screening disorders are a great example of that, where there is that action that a physician can take. But in a situation of variant of uncertain significance where the enzyme levels come a little off limits when in the newborn screening, the physician has to apply that right clinical judgment. And if it is not going to do any harm, I think it’s important and I know that the physicians will take that right step, that they can support the family with these kind of different options. But then there are these disorders where you just can’t do anything, and it’s a very hard thing for the physician. Not only are they going to try and take care of that child, the family, but also giving advice for if the family wants to have more children and you’re doing all of that at the same time, unfortunately, when the child is not doing well probably, and it becomes such a personal decision for the family, and this is where compassion plays a huge role in doing that education piece and how you apply that knowledge. In between these metabolic disorders and very severe conditions like these conditions where there is evidence now that you know certain things can be done. I think epilepsy is a good example of that, you know some states approving certain — and then some states not doing it, and families just really craving for the ability to have access so that they can see it help their child. I think that becomes a very difficult decision, so compassion at all levels, but different categories of it.
GARLAPOW: I think since moving into this industry, my understanding of the clinical shape of hope has really become quite complex.
HEGDE: I think after doing 20-25 years of genetic testing, and finally what goes to the counselor and the physician who is giving advice to the patient, it becomes such a personal decision. You know, no two individuals are going to look at the result the same way because it is not just that definition of what that clinical condition is. There’s a whole social aspect of it, there is a family aspect of it, and that has to be taken into consideration.
GARLAPOW: Oh absolutely, absolutely. So variant classification is, it’s hard. Why can’t artificial intelligence save us? Why can’t we just write some excellent code and, you know, let the robots save us all? Why isn’t that an option? Why is that actually a really dangerous choice, to rely exclusively on artificial intelligence?
HEGDE: The last two years have been interesting, right, with AI, that there’s so much hype about AI. Definitely AI is going to be very useful going forward, so it’s not like we’re all saying that, “Oh this is very, very dangerous and don’t do it,” but the human aspect of that cannot be removed, because variant classification is a science but also an art, and there is that human piece of it which you just cannot remove. A great example of that is our databases. Today with ClinVar or with so many other databases, you have the HGMD database, the LOBD, and a multiple number of databases which operate in different regions of the world, let’s not forget that we are spread all across the globe, and certain variants are common in certain regions which you may not even know of today and we are classifying them as pathogenic. This is a very dangerous situation. Until such time, we have sequenced (maybe I’m exaggerating here) maybe half of the world’s population, and make databases so powerful that we can actually interpret variants. And I have learned a lot since I have come to Perkin-Elmer. At Emory Genetics, I was really doing only genetic testing for cytogenetics, biochemical, and molecular genetics. What I added when I came to Perkin-Elmer was newborn screening, because Perkin-Elmer is a leader in newborn screening. I have learned so much. I have learned newborn screening is not an assay, it’s a system. It’s an entire system that has to work in such an efficient fashion, that you cannot rely on something by saying that AI can save us all. It just will not work. That’s one aspect of it, and also Perkin-Elmer is a manufacturer of a variety of different in automation and assays. And how much time is spent in R&D, and doing due diligence in coming up with these assays, which eventually go through a rigorous FDA approval. We have seen what has happened in the COVID era with the emergency approvals that have been given. So, it’s a complex environment, and I don’t think AI is the final solution of it. The one thing which we do not even know today is the outcome. The outcome of what happens at the other end, very rarely that feedback is given. Newborn screening, even today with newborn screening, the last publication I think came out sort of looking at the outcomes was somewhere in late 80’s. I was on a committee recently where we were asking this question and none of us could come up with a paper that we could all rely on to look at. So, this AI, the final “go all be all.” I don’t think so. And I think it is dangerous, at least today. When we will get there? I don’t know, it depends on how we all come together to at least have a common approach, maybe not a consensus. So today it is not ready but it’s definitely a tool amongst the many tools a clinical laboratory uses to do variant classification.
GARLAPOW: I couldn’t agree more, Madhuri. Thank you so much. So, some health insurance companies have started supporting reimbursement policies that only reimburse if the report is positive for a pathogenic or likely pathogenic variant. What effects could this have on reports, and what can laboratories do to combat this?
HEGDE: So, first of all, I think it’s really unfortunate that insurance companies are looking at clinical reports like this because and this is again goes back to education, there is so much more that goes into interpreting a variant and you do not want a laboratory to interpret a variant pathogenic just to get reimbursed. It will cause chaos, because there’s just going to be so much over-interpretation that is going to happen. Also, understanding that the effect of a variant on the entire, not just the management or the care for the patient, it is about the entire family. Many times, the additional family members need to be tested to understand segregation and actually help with variant interpretation. There is just so many aspects of that that the insurance companies are ignoring with these kind of policies. So what labs can do I think mainly it’s publishing data. I think there is just no replacement for that, as busy as we get in our routine, day-to-day reporting of cases is very, very important to publish data. That data has to be published in journals which can be broadly accessible. And education. I think education is the most critical need today and education done by advocacy groups. They are very active. I think one of the sort of great outcomes of the new era of genetics is these advocacy groups really campaigning for the disorders of their interest. They have brought this focus on the families and have gone and had discussions with the insurance companies. So I would say insurance companies need to reconsider this, but clinical laboratories need to work very closely with the advocacy groups, with organizations such as ACMG, ASHG, AMP to go and campaign and do this educational side of it.
GARLAPOW: Thank you, Madhuri. Yes. Yes, that’s my response. So for clinicians and patients not as versed in this space, what do you hope their top take-home message is from this conversation we’re having?
HEGDE: I think I would say about three important sort of take-home messages here. The first one is that variant classification does require human intervention, and it is not something you can rely on just looking at databases. When you get a clinical report, talk to the laboratory, ask questions. What testing was done? What are the missing holes? Are there proper disclaimers or not? Absolutely critical when taking a clinical report and applying it to patient care. The number two thing is that the research aspect side of variant classification is, actively contribute back to the community as well, by working or collaborating with the lab so that they can accumulate the data and publish it. It is difficult for clinical labs to get IRB consents or go and approach the patient. The physician and the genetic counselors are that link in between, and this is really important. And the third thing is giving the outcome information is also very, very important, so that we can fine-tune this art and science of variant classification and move forward with bringing in new technologies for clinical care.
GARLAPOW: Wow. Okay, that leads perfectly into my last question for you Madhuri. This is The Rising Tide podcast. A rising tide lifts all boats, improves quality, improves accuracy. We’ve covered tons of ground here Madhuri. How can we raise the tide?
HEGDE: To raise the tide, I think we are heading in that direction definitely, but what do we need to do is collaboration. Today, no one can work in a silo. Collaboration, collaboration, collaboration, and education. Like this podcast, is absolutely critical, and working with all agencies across the board. As I just mentioned, the advocacy groups, the organizations, and even organizations such as College of American Pathologists which are putting out this proficiency testing. Looking at the accuracy, and maintaining that accuracy is so critical. So working together collaboratively and participating in proficiency testing, letting that information come in and go out as well is absolutely critical.
GARLAPOW: Oh, completely. Madhuri, I am so grateful for your sharing just even snippets of your expertise. I could speak with you for hours about this, so maybe I’ll have to have you back on the podcast in the future so that we can get back into this, but I am so grateful for this and I’m so excited to share this episode with the broader world, to get some of these messages across for everyone.
If you loved this episode of The Rising Tide podcast, please support us on Patreon and select us on Amazon Smile. The Center for Genomic Interpretation can help your organization identify high quality clinical genetics and genomics partners or can help a laboratory identify areas for quality improvement. Through ELEVATEGENETICS, we engage with laboratories, encourage genetic and genomic test validity, and assess test efficacy. Find us at clarifygenetics.org and genomicinterpretation.org.
For our listeners following along on YouTube, remember to hit subscribe and to activate alerts for when we post new episodes. You don’t want to miss any of these important conversations with leading experts. Together we can raise the tide.
Narrated by Dr. Megan Garlapow
Produced and edited by Kathryn Mraz, Hunter Giles, and Brynlee Buhler