Unlocking the diagnostic potential of whole exome sequencing: The role of high-quality interpretation

Welcome to today’s educational webinar titled Unlocking the Diagnostic Potential of Whole Exome Sequencing, the Role of High-Quality Interpretation. The webinar is brought to you by Blueprint Genetics, a genetic knowledge company committed to provide an innovative approach to genetic testing and to ensure accurate and confidence in your clinical practice. My name is Taina Vuopio and I have the privilege of hosting today’s webinar. Please submit any questions you may have in the QA box. You can submit them throughout the webinar, and we will answer as many as possible at the end. Oh, we are excited to have Doctor Kirsty Wells as a speaker today. Doctor Kirsty Wells is a Clinical Interpretation team Manager at Blueprint Genetics who specializes in whole exome sequencing, interpretation and reporting. She has a background of both research and diagnostics. Before joining Blueprints in 2018, Doctor Wells completed both PhD and postdoctoral research fellowship and undertook in depth training in genetic diagnostics in EU KS National Health Service. Thanks for being here today, Kirsty, please. Thank you, Taina. Thanks very much for the nice introduction. Good afternoon or good morning, everybody, depending on where you are. Thanks so much for joining us today. So as Taina said, my name is, is Kirsty Wells and, and, and it’s a real pleasure to be here again today to talk to you about whole exome sequencing interpretation. Because of Blueprint Genetics, we really believe that without high-quality interpretation, genetic testing can’t deliver its full value. OK, so let’s start with a quick introduction to our company. At Blueprint Genetics, we deliver high-quality NGS based genetic testing to the global clinical community. We do that through a range of NGS based test types including gene panels of course in 14 medical specialties, also whole exome sequencing, single gene and variant specific testing and also screening. And the results are ready in approximately 4 weeks or it’s also possible to get results in in just two weeks now with our new express service. But whether we’re talking about NGS panels, single gene, or whole exome sequencing, the real value of all of these tests comes not just from the technical performance of the test, but from the skilled interpretation of the results. And this is arguably the most important piece. So that’s what I’m going to talk about today. I’ll cover some of the key lessons that we’ve learned at Blueprint Genetics over the years about what drives high-quality interpretation and specifically how we’re applying this to our Wes interpretation strategy. Because whole exome sequencing or Wes as it’s affectionately known, does have some interpretation challenges. And the main one is the sheer volume of data compared to more targeted tests such as gene panels. In a typical Wes, the patient has around 35,000 SMVS to deal with. So that’s before we even get to the CNVS or the mitochondrial variants. So that’s a lot of a lot of variants to deal with. So, first, effective variant interpretation for Wes really depends on having a good tools and a good robust process to deal with all of these variants. So, a good way to filter them, prioritize them and accurately annotate all of these called variants so that we can only look at the potentially relevant ones. Because of course we can’t look at all 35,000 variants, but we want to make sure we don’t miss anything. So that’s challenge number one for WES interpretation. How, how to pull out a manageable number of relevant variants, including of course, the diagnostic ones from these thousands of variants called, but without missing anything in the process. And then once you’ve narrowed down to a manageable number of variants for review, the next challenge is how to get from that to the variants that you want to report. So, we’re talking typically one or two clinically relevant variants and, and, and there are added layers of complexity with whole exome sequencing here as well. First of all, we find that patients referred for comprehensive testing like whole exome sequencing, have often had previously inconclusive genetic testing. And our experience shows us that that’s often because the patient’s phenotype is atypical of a known disorder. And so that has meant that that for the clinician, perhaps it’s been difficult to choose a more targeted test for that patient. So, we see that very often in whole exome sequencing cases that the patient’s phenotype is not quite what you might expect for the molecular diagnosis that they eventually, eventually get. And then we also find that patients that have been referred for whole active sequencing with us may have variants that were actually missed by earlier testing, often because the diagnosis actually lies in a very recently characterized disease gene or because the variant or variants concerned were inherent, inherently difficult to interpret. So, these are important things to consider when it comes to a Wes interpretation strategy that the patient’s phenotype is often atypical. Diagnosis can often be in newly characterized disease genes and the variants; the reportable variants can be challenging to interpret in some way. OK, so with these interpretation challenges in mind, I’d like to now take you through four of the most important elements of our Wes interpretation strategy, illustrated with some interesting and of course anonymized cases from the Blueprint Genetics Archive. OK, so let’s start first with the importance of keeping gene disease associations current and let’s jump straight into a case. This case involved a young adult who had died suddenly and had dilated cardiomyopathy, and a trio Wes was ordered from us which included data from both healthy parents. On the Wes, we identified compound heterozygous variants in the NRAP gene which were very interesting because both were rare. They were in compound heterozygous state, and one was a frameshift inherited from one parent and the other was a missense variant inherited from the other parent. So what’s about this gene NRAP? Well, looking more closely at this gene, at that time when the case was interpreted, there was no gene disease association for NRAP in the places that we normally look to get gene disease association information. So for example, on OMIM, on Clingen, there was there was no gene disease association listed there. So let’s go back a little bit now to talk about this concept of gene disease association. What does it actually mean? Well, what it means is that basically that there’s a relationship between a particular gene and a disease. And when you’re interpreting and classifying variants, it’s vital to take into account the strength of the evidence that supports the relationship with disease for that particular gene. So at Blueprint Genetics, our gene disease association classification scheme is based on the Clingen framework. And so there are several different classifications for gene disease associations, and our W analysis aims to focus mainly on those gene disease relationships that are classified as moderate or above. So everything in green on the slide because any gene disease association classifications below that, so emerging limited or below they are, they are not considered clinically relevant gene disease associations at that time. So they don’t have sufficient evidence to support them. That means that variants in those genes are always going to be of unknown relevance if there is no, if there is not enough evidence to support the association with disease for that gene. So we don’t, we don’t generally report variance in those genes in our exomes because we want to concentrate on clinically relevant findings in our exomes. So we’re looking at gene disease associations classified as moderate or above. But of course, this means that that having up to date and correct gene disease associations in whatever system you’re using is really crucial. Otherwise, it can lead to missed relevant variants. And what we find is that that GD as on databases are not always current. So to illustrate this, let’s go back to our NRAP gene. I said there was no gene disease association for this gene on places like OMIM, but in fact there had been a publication that came out well before this case was interpreted. And this publication reported 11 unrelated PRO bands with dilated cardiomyopathy with biallelic or two NRAP variants of unknown phase. And also there was found to be significant enrichment of biallelic NRAP variants in patients with DCM. And also a CRISPR knockout of zebra fish was found to have a cardiac phenotype. Based on this evidence and using our classification scheme, we consider that the the gene disease association between the NRAP gene and dilated cardiomyopathy is classified as a strong association. So that means we can classify variants in this gene as likely pathogenic and pathogenic. And in fact, based on this gene disease association on the properties of the variants themselves and also reports of these variants in other patients in the literature, we classified both of the NRAP variants in this patient as pathogenic. And so therefore this patient was compound heterozygous for pathogenic variants in NRAP, which explains their dilated cardiomyopathy. So what’s really important to note here is that if this gene disease association for NRAP was missing in our system, these variants could be missed. So especially if you’re using AI driven variant prioritization, it’s, it’s really important to make sure that your gene disease associations in your system are up to date and correct to avoid missing relevant variants in, in genes that have been associated with disease relatively recently, for example. And we, we recently tested a number of whole genome sequencing interpretation tools that are available on the market. And this NRAP pair was not prioritized by any of them due to, to missing gene disease association. So this, this illustrates how important it is to, to keep those GD as up to date in whatever, whatever system or tools you’re using. OK. So therefore, this patient received or this patient’s family received a molecular diagnosis of NRAP related autosomal recessive DCM for this unfortunately deceased patient. But this facilitates genetic counselling for the family and also risk evaluation, particularly for any siblings of the patient who, who would also be at 25% risk of inheriting both of these variants in compound heterozygous state and, and being affected. So this, this diagnosis, molecular diagnosis was, was achieved thanks to keeping gene disease associations in the system up to date and, and current. And this is not an easy task because there are new gene disease associations being published all the time. So this is a big challenge for us and other labs as well. But at Blueprint Genetics, we, we tackle this by first of all really leveraging our internally curated gene disease association knowledge. So we have a wealth of, of knowledge gathered over the years and recently on evidence supporting gene disease associations. And we, we do our best to make sure that that information is in our system and that we base our gene disease associations on that and not relying solely on external databases for these GD as because, as, as we’ve shown, they’re not always up to date, understandably, because the literature moves fast and we continually monitor and curate the new literature and try to make sure new gene disease associations are updated in our systems as frequently as possible. And yeah, this is really vital because we, we actually estimate that there are at least 20 new gene disease associations published every month. So all of these shown on the slide I pulled from one month during the summer and there there were more as well. So there’s a lot of new literature all the time. So it’s vital for effective variant prioritisation and interpretation that that we keep up to date with this as as best as we can. OK, so moving on now to the next interpretation principle which calls for us to prioritise effectively. And let’s start again with a case. This patient was an adult who had severe intellectual disability, absent speech, short stature, neuropathy, ataxia and dystonia and also hearing loss. So, so quite a a complex phenotype and testing elsewhere had apparently been negative, although we we don’t know what that that testing was a pro band only WES analysis was ordered and the Wes identified a mis sense variant in the Morsi 2 gene. This variant was predicted to be deleterious by the insilico tools that we use. It was very rare, so absent in Nomad, but it wasn’t found in any patients in the literature or on disease related variation databases or in our in-house data previously. So not much to say about this variant other than it’s very rare and it’s predicted to be deleterious by in CIVICO tools. So this variant is VUS, but what was interesting too was about this variant is that it was the number one prioritized variant by our variant prioritization tool, Alva. So what is Alva? Alva is, is our in house developed variant prioritization tool and, and what it does is it gives each variant A pathogenicity prediction score between zero and one, with 0 being a low probability of pathogenicity and one being a higher probability. The score is based on very many variables, more than 50 and these are all variant properties. And it works on single nucleotide variants and and short in Dells. It’s important to note that it it prioritizes variants for us, it doesn’t actually classify them. So. So the geneticists do that still, but Alvar prioritizes the variants for us, and the score that it gives us is independent from the case and the phenotype of the patient. So the Alvar score for the variant should remain the same regardless of the patient the variant is found in. And we find that Alvar works very well. This slide compares how well different in silico tools can predict whether a genetic variant is deleterious in a large data set of of variants. So so the metric used along the the X axis. This is a score from zero to 1 where one means the tool is perfectly accurate in predicting deleteriousness, and Alva has an extremely high positive predictive value at naughty .998, which means. What it actually means in practice is that it correctly distinguishes deleterious variants from benign ones almost every time, and we found that it outperformed all the 16 other tools tested. That means that we can confidently use ALVAR to help in prioritising variants for us in West cases based on strong variant properties. Let’s go back to our case with the more C2 variant. This gene, more C2 is a bit of a tricky gene in terms of phenotype matching because the phenotype associated with this gene is is really variable. There are very many different clinical features, different phenotypes and and perhaps a bit of a spectrum going on on OMIM. There are actually 2 phenotypes annotated there. For more C21 is an annotated as axonal CMT and then there’s another phenotype which is a more early onset syndromic disorder characterized by developmental delay, impaired growth, dysmorphic features and also axonal neuropathy. So there’s a lot. So the phenotype matching with a gene like this can be a bit tricky. But here if we compare the phenotypes in our patient with those reported in association with more more C2 related disease in general, we do have plenty of overlap there. So Alva alerted us to this variant straight away and then the geneticist determines that there was some phenotype overlap meaning that the variant was reportable and it was reported as a VUS. And then subsequent parental testing did not detect the variant in the unaffected parents, which is consistent with de Novo occur occurrence in this in the patient. So based on our classification scheme, this variant was upgraded to to likely pathogenic. So in summary, whole exome sequencing and and subsequent parental testing for this patient identified a de Novo likely pathogenic variant in more C2, which confirmed a diagnosis of autosomal dominant more C2 related disease. And and this this was thanks to intelligent variant prioritization. Alva prioritizes variant, so we were able to focus in on it a lot more quickly than we would without this tool. In West cases, there are a lot of variants to go through, so it’s really helpful to have a variant prioritisation tool that we can we can rely on to help us get to the diagnosis more quickly. Because this particular variant, it was just a VUS, no literature, no previous observations in patients, no functional studies. But Alvar prioritised it as #1 for us. And indeed, it did end up being the diagnostic variant. So yes, this kind of variant prioritisation is really valuable in WES cases at Blueprint Genetics. OK, so now let’s move on to the the third interpretation principle, which is connecting genotype and phenotype blueprint genetics. We’ve long followed what we call a genotype first approach in in our in our tests. So what does this mean? Well, if you’re thinking about whole exome sequencing, what it means is that all of the genes on the target are sequenced. We don’t remove any genes from the analysis at this first stage based on lack of match with the patient’s phenotype. So everything is sequenced and then the variants are filtered and prioritized based on their properties. We don’t do any phenotype-based filtering or prioritisation in our whole exome sequencing cases. So the geneticists see, see all the variants regardless of phenotype match and then the genetically strong variants are assessed by the geneticist at this last stage for phenotype overlap. And the variants are reported if they could explain all or part of the patient’s phenotype. And this approach makes a difference in a number of different scenarios. First of all, it can be very powerful for patients with atypical presentations, which I’ve said, I’ve said a couple of times that that in whole exome sequencing cases, this is something that we see quite a lot that patients don’t, don’t always have the clinical features that you would expect for known, known disorders. So genotype first really makes a difference here because it doesn’t, it doesn’t require the phenotype match to be perfect for us to identify a variant and and consider it for reporting. Similarly, it can be helpful for very young patients where the full clinical spectrum of a disease may not be, may not be visible yet. So for those patients it can be very difficult to do the phenotype matching. So it helps to to be driven first of all by genotype and then to think carefully whether the phenotype could fit with what we do have reported in the patient. Similarly, cases where clinical information is sparse or incomplete and the phenotype matching can be difficult for that reason. And also it can be helpful in if the diagnosis ends up being in a gene with a very new gene disease association where again, the full kind of clinical features of the disease may not be known yet. Genotype first makes it possible to identify those kinds of variants as well. So very many scenarios where following a genotype first approach really helps us identify relevant variants. So let’s take a look at the case now to illustrate this. This was another pro band only West case. The patient was a child with mild global developmental delay, febrile seizures, and brain MRI abnormalities. The West identified A heterozygous missense variant in the ZEB 2 gene. Again, not an awful lot to say about this variant. It kind of had a mid range rebel score, which is the in silico tool that we use. It was absent in in Nomad and was not found in the literature or on databases or in house previously. Again though, this variant came to our attention immediately because it was prioritized as the number one heterozygous variant in a gene associated with autosomal dominant disorder by by our prioritisation tool, Alva. Looking at the disorder associated with this gene, ZEB 2, Pathogenic variants in ZEB 2 are most commonly known to cause an autosomal dominant multiple congenital anomaly syndrome called Moet Wilson syndrome. This is typically characterized by a distinct facial phenotype, often severe intellectual disability, absent speech, epilepsy, Hirschsprung disease, congenital half heart defects and also growth restriction and microcephaly. So looking at these clinical features, I’m sure you’ll agree this is not a good match, if any, with what is reported in our current patient. But because this variant was genetically strong and it was prioritised as the top variant by our prioritisation tool, the geneticist is not going to give up as easily as that. And she did some really careful digging just to make sure that there really is no fit between ZEP 2 and what we see in our patient. And actually what she found in the literature was that three patients had been reported with de Novo missense variants in this gene with a mild form of of Moat Wilson syndrome. All of these variants were in a highly conserved zinc finger domain, which interestingly our current patients variant was also located in in that domain. And these patients had subtle facial features that could go missed unnoticed, more moderate intellectual disability and none of the congenital anomalies that are typically reported in ZEB 2 related disease. So because of this literature, we decided that the variant was definitely worth reporting and the variant was reported as a VUS. And then actually family member testing subsequently showed that the variant is is de Novo in the index patient. So we we believe that this makes it more likely that this variant is the diagnosis in this patient. So to summarize this case then a relevant variant in ZEP 2 was reported in this patient, which is which is a very possible molecular diagnosis. And this now facilitates genetic counselling and and risk evaluation. And the point here is that if you’re filtering according to phenotype, a variant like this would likely be missed because there was not a very good match at all with the features that are most commonly associated with ZEB 2 related disease. So in where’s that blueprint genetics. For this reason, we don’t filter according to phenotype and and our in-house Alvar tool strengthens our genotype LED interpretation, enabling us to identify findings like this in patients like this one with an atypical phenotype. OK, so let’s move on now to the final key elements of of of where’s interpretation. And that is the application of human expertise. And I’m probably very biased, but this is this is my favorite, this is my favorite 1. So let’s again jump straight into a case. This patient was a child with mixed axonal and demyelinating sensory motor neuropathy. And this child had a similarly affected parent. And actually testing elsewhere had already identified an interesting variant in the PDE 8B gene, a nonsense variant, which had been classified by the other lab as pathogenic. But the customer wants to continue looking and ordered from us a duo Wes, so including data from the affected index and they’re also affected parent. So first of all, the Wes analysis did identify and prioritize the previously identified nonsense variant in PDEHB. So of course, we take a careful look at that one. This variant was rare, so absent in nomads. So that’s consistent with the role in autosomal dominant disease. This variant is a nonsense variant predicted to lead to a loss of function. And actually loss of function has been reported as a mechanism of disease in this gene. And then taking a look at the phenotype associated with PDE 8B, this, this gene is associated with an autosomal dominant disorder called striatal degeneration. So at first glance, this is the diagnosis for this patient, right? Actually, no. Of course, that’s why I’ve chosen this example. The plot thickens with this one. So first of all, we noticed that this variant was not inherited from the affected parent. That was the first red flag for this variant. It did not segregate with disease. Also actually looking closely at the disease associated with PDE 8B, it the phenotype match was poor with what we’re seeing in our current patient. This striatal degeneration is primarily a movement disorder which which our patient does not have. They have neuropathy which is which is different. So the phenotype match was poor. And also looking more closely at this variant, even though it’s predicted to lead to a loss of function, we noticed that this variant also was predicted by in silico tools to have a splice effect, which this is a bit speculative, but it may lead to an in frame change and not actually cause a loss of function. So, so, so this is speculative, but together with the other evidence about it not segregating with disease and the phenotype match being poor, we considered that this variant was potentially a false lead in this patient and we classified it as a as a VUS. So the geneticist on the case carried on looking at what else this patient had and they found an interesting heterozygous missense variant in in the SLC 12A6 gene. This time the variant was inherited from the affected parent. The variant was genetically strong is absent in Nomad predicted deleterious spine silico tools and also interestingly there had been a likely pathogenic variant reported in a nearby codon suggesting that this this region of the protein is functionally important. So this variant was reported instead of the PDE A/B variant and it was reported as a strong VUS. But we believe that this variant is is the more likely diagnosis for this patient than the PDE 8B variant. So this case is is a nice example of of when a human touch is really needed to look more closely at the findings and and consider it in in the case specific context. And there, there are many elements of of interpretation where a human, human kind of intervention is really needed to get the right result. So for example, when you have multiple findings in one patient, when you need to assess a gene disease association and gather the evidence for that, interpreting the literature, looking at contradictory literature or functional studies, interpreting difficult variants, which we have every day, assessing phenotype match and, and yeah, as in this case, considering the case specific context. So This is why we have a really large team of very skilled geneticists at Blueprint Genetics who do this kind of thing every day. Leverage the powerful tools, but have a deep awareness of of the limitations of them and how to exploit those tools to get the best result and the right answer for the patient. OK, so to summarize what I’ve shown you today, I’ve talked about four of the key elements to high-quality WESinterpretation at Blueprint Genetics. First of all, I talked about the importance of having up to date gene disease associations on the system. Also, the value of having intelligent variant prioritization of all of those variants that you get in. Where’s the power of a genotype first approach and also the the power of having human expertise in your workflow? OK, so it just remains for me to acknowledge my many amazing and wonderful colleagues in the clinical interpretation team at Blueprint Genetics who work passionately every day to unlock answers for patients and families. And thank you to you for listening. Happy now to take any questions. Thanks very much. Thanks, Kirsty. Excellent presentation. I, I really like the patient cases you gave. They tie everything together. Yes, we have actually quite many questions. OK. The first one is how exactly do you use Alva in this interpretation? Yeah, good question. So we use it specifically to prioritize the rare heterozygous variants for us in WES analysis. Going through all of the rare heterozygous variants associated with autosomal dominant disorders is really the biggest burden for the geneticist and it can take a long time. So it’s really helpful to have a tool that can rank those for us according to variant properties to help us get to the, the the relevant variants more, more quickly rather than having to look through a very long list of variants first. OK. So we then again I’m going back to our which filter do you use first, our L frequency or our score? We use both of those filters actually. So we combine, we combine frequency and Alvar score to really bring the the most relevant rare variants up to the top for us. OK. Then the next one, what is the rate of ALVAR predicting a potential variant among say 100 cases? Could you say this is always 100%? If not, what factor could cause this gap? Yeah, really great question. I’m not sure we have any numbers on that, but we have. We performed a validation of of our ALVAR tool and we find that it’s highly specific and sensitive in in identifying the relevant and reportable variance. The things that can make a difference in in how accurate Alva is. So Alva it works very well for for missense variants and in Dells it doesn’t always work so well for other variant types like intronic variants and frameshift variants. So we have to be a bit careful with those when, when when applying the Alva score. So, so, so the variant type can can affect how how reliable alpha is for us. OK, the next one, what if a variant is genetically strong, but there is no phenotype overlap found? Yes. So in in whole exome sequencing, if there really is no match with the patient’s phenotype, then then we don’t we don’t report because in where’s we want to answer the clinical question. The only exception to that is is in the context of of ACMG secondary findings and if and if there is consent for that, but otherwise no phenotype match, no, no reporting. But then having said that, if as I was explaining with the ZEB 2 case, if if we have a genetically strong variant and there it doesn’t initially appear to be a phenotype match, then we do work hard to really make sure that there is no match. We look carefully at the literature that there aren’t some parts of the phenotype that could be explained. But otherwise, if there’s really no match, we don’t report. For example, if it’s a pathogenic variant for an adult-onset disorder in a child, those kinds of findings are not reported in our ways. OK. The next one, what kind of evidence is considered when classifying a gene disease association? Yeah, really great question. So our gene disease association scheme is based on the Clingen, Clingen scheme, which is very kind of robust and well established. The type of the type of evidence that we consider primarily we consider published pro bands with variants in the gene of interest and how many of those there are. And also of course the properties of the variants that are reported in in those pro bands. And we also consider additional evidence such as functional studies supporting the gene disease relationship, mouse models, segregation data, whether the variant segregates in in families, whether the variants are de Novo. And primarily we lean on the literature for all of that information, but also, we we can bring in data from other sources, such as our in-house data or if there’s good data on a database like Clinvar for instance, we can also use that. So yeah, we gather all of this evidence and we apply points for the evidence that we gather in order to meet, meet the classification, the GDA. That’s how we do it. OK, the next one is also. This is a long 1. You mentioned that Alvar performs genotype first prioritization independently of phenotype. How do you prevent clinically relevant variants from being deprioritized when the pathogenic mechanism is disease specific? For instance, reduce penetrance, a typical inheritance or variants whose significance becomes apparent only after phenotype driven reanalysis. Well, I’m not sure I fully understand the question, but I would, I would expect our Alva tool to still prioritise those variants because if they are, if they are genetically strong, then then they will be prioritised by our Alva Alvar tool. And then we assess things like whether the phenotype could match and whether there is incomplete penetrance and, and, and all of those things. Not sure I answered that very well. I don’t know if the questioner wants to elaborate more or message me afterwards to discuss. Happy to do that. OK. And then the next one, at which which stages of this interpretation does Blueprint currently use AI driven variant prioritization and which decisions always require review by a clinical genetist or molecular genetist? Yeah, First of all, I wouldn’t, I wouldn’t say that our ALVAR tool is AI assisted. So we’re not using any AI in our ways at the moment. But yeah, ALVAR is a prioritization tool and we use that specifically at the step where we are assessing the rare heterozygous variants and the types of findings that are reviewed by a second geneticist are or if a variant is is, if the case is particularly complex, if a variant has been seen for the first time by us by and it’s classified as likely pathogenic or pathogenic, There are a number of scenarios where that there has been some elements of interpretation in the final report that needs a second, second check. So yeah, we have really tight safeguards on the the kind of findings that need need a second eye, OK, then the next one when the genotype looks convincing, but the phenotype is are typical, how do you exclude a dual diagnosis? Yes, good question. Because dual diagnosis and sometimes even more is something that we we do see not frequently, but it does come up. The answer to that is we keep, we keep looking. So if we find a diagnosis, molecular diagnosis that that explains that doesn’t explain the whole phenotype, then we keep looking at the variance to see if there’s something else which could explain the rest of the phenotype. So if we find something diagnostic, we don’t we don’t generally stop there if it doesn’t explain the whole phenotype. OK, then the next one. Do you follow the Jennifer first approach for panels? Yes, yeah, we do. We follow this approach for or for all of our most of our test types where it’s relevant. So for panels, the same principles apply that we don’t do any phenotype-based filtering in panels and we report according to whether the variant is genetically strong and there is some phenotype overlap when we’re talking about the primary findings in the report. It’s useful for panels. I would say it’s especially useful for where’s because as I’ve said with where’s we often see it that the patient has an atypical phenotype. With panels that happens less often because those patients often have clinical features suggestive of a specific genetic disorder. And, and whereas that’s not always the case. But yes, with panels it can be very, very powerful as well as in where’s. OK then how confident can we be in ACMG classification for a rare variant in a Pakistani or South Asian patient when anxiously matched population data are limited? Yeah. Could a variant appear so species simply because South Asian are underrepresented in reference databases? Yeah, absolutely. And this is something that we meet all the time that, that if, if a population is underrepresented in Nomad, then we have to be really careful with the classifications. Something that can help us a bit with those kind of situations is our internal data. So we have, we have a lot of internal data on previous patients and, and variants that have been seen in them. So it, it sometimes that might help us to, to work out whether it’s just a common variant in that population that’s not in nomad or whether it could be disease causing. But we yeah, with those sorts of situations, we tend to be a bit more cautious with the classifications, especially, especially if if, if we have only seen the variant as homozygous in inpatients. OK. Hey, thank you, Kirsty. Thank you. Great questions, excellent presentation and answering all these questions. So then the recording of this webinar will be made available, and we will send it through e-mail as soon as it’s it’s ready. So thank you again for everybody and I and I hope a good rest of the week for everyone. Thank you, Thank you everyone. Bye, bye, bye.