Good afternoon. I'm Max Mönke. I'm the CEO of the American College of Medical Genetics and Genomics, and I wanted to welcome you to what I consider one of the most exciting webinars that the college has to offer. I mean, it's both because of the topic, and I love this topic, next-generation sequencing can't be more topical than that, and then the panel here. But I wanted to just give just a few remarks to those who participate. We have over 400 people signed up for this webinar, probably the largest, if not the largest that the college has had in its history, so welcome. I wanted to welcome all of you, but those of you who have, who are thinking about medical students, PhD graduate students, would they want to go into the field of medical genetics, please contact Jane Radford afterwards, get my email address, and we will talk, because I've been in the field for 30 years. I've worked in clinical genetics and molecular genetics and in cytogenetics, and for 30 years, it has been the most rewarding, meaningful career that I ever could have imagined coming from a small town in northern Germany. So I welcome you to this wonderful, exciting webinar, and please contact me, and don't be shy. Have a good work, good workshop here. Thank you very much, Max. My name is Janelle Hodge, and I'm really, really thrilled to be able to help bring to you this webinar, which is Essential Fundamentals, Next Generation Sequencing from the Beginning. And this is actually the end portion of a short course, which involves some amazing educators, Elaine Martis, Susan Hsiao, and Mark Ewald, and we're going to continue on to present to you today. So in terms of disclosures, myself and Dr. Ewald have nothing to disclose. Dr. Martis does do consults for CATUS Pharmaceuticals NV, PACT Pharma LLC, Interpreter LLC. She does have stocks and ownership and her interest in QIAGEN, and is part of the board of directors and a supervisory board of QIAGEN. Dr. Hsiao consults for Bristol-Myers Squibb and Lokso Oncology, and also has some external grant research support from Bristol-Myers Squibb. So what really brought this whole short course together was there was a survey sent out on the CytogenX listserv by Hutton Kearney from Mayo Clinic, and she wanted to ask participants, you know, what areas are you really feeling you need more educational content? What are you comfortable with? What are you not comfortable with? And about 130 people responded, so a pretty good number of people, and this was across the board, clinicians, lab directors, technologists, and what happened was is that 63%, which is a big percentage, said they have significant educational needs in NGS. And I know that what happens is there's lots of NGS education out there, but most of it seems to start from the intermediate level on. So the idea of this course was to say, okay, can we present NGS at the most fundamental level? Definitions, you know, what are the basic principles, the methods, what kind of platforms are used, what the applications are, and then when you see a file format, you know, what is a VCF? What is it composed of? What step do you use it in your bioinformatics pipeline? What QC metrics are important? How do you validate one of these things? How do you look at and identify your technical artifacts? How do you troubleshoot them and more? And so we have quite a bit of hours, four plus hours of prerecorded content that covers this information. And the idea of this session today is to apply that knowledge in a case-based format and also to help answer your questions. So hopefully you enjoy it. I guarantee these are some amazing speakers. If you haven't gotten to look at those talks, they're full of really helpful information. Even if you think you know NGS, they fill so many holes of little principles. Oh, that's how it works. That's why it's that way. So I really encourage you to go back and watch those talks if you haven't. We also online provided you with this document, which is a definition that hopefully can help you now and in the future if you ever want to remember what something is specifically. Okay, so again, my name is Janelle Hodge. I'm an associate professor and a director in cytogenetics and molecular genetics at Indiana University. My prerecorded talk was the Next Generation Sequencing Education Gap. We also have Elaine Martis, who is co-executive director at the Institute of Genomic Medicine at Nationwide and the Nationwide Foundation Endowed Chair of Genomic Medicine at Nationwide Children's Hospital. She has two prerecorded talks. One is Next Generation Sequencing Technologies and the other is What's New in NGS. We also then have Susan Hsiao and she's an assistant professor at Columbia University and her prerecorded talk was Basic Principles in Bioinformatics for Identification and Interpretation of Genomic Alterations. We round out that dream team with Mark Ewald, who's an assistant professor and he's also an associate director of the Colorado Molecular Correlates Laboratory and director of the Molecular Genetics Pathology Fellowship at the University of Colorado and Schutz Medical Campus. His prerecorded lecture was Approaches to NGS Assay Design and Validation. So what this live webinar is going to consist of, I'm doing this little introduction now, but we're going to then jump into the case-based application and you will see four cases presented and after each case, we will have a Q&A portion. So please be submitting questions during this for the cases and we'll try to address case-based questions at the end of each. If there are additional questions that are not necessarily specific to those cases, the idea then is after we finish that portion, we'll have some time to do additional Q&A and I have a few pre-submitted questions from before this webinar that I have here to answer and as well as new questions coming along with this webinar. Very important, just logistics. There's a few ways to submit questions but really I'm only going to be able to monitor the Q&A button. So if you look at, usually it's at the bottom, sometimes the top of your screen, you'll see this little Q&A button, click on it and once you click on it, you can watch any questions others are submitting as well as submit your own questions. Also, if you want to, if you like a question, you think, yeah, I really want that one answered, upvote it. So you'll see next to the question, once you have this box open, this little thumbs up, right? Click on it and when I see more and more numbers climbing, this jumps it up the list and indicates to me that's a more important question. We are not going to be able to use the chat function so if you try to submit a question in chat, I won't see it. We have ACMG staff that will try to then guide you to submit it through Q&A but if you're having technical problems, please go into chat and ACMG staff will help you there with that. Also, there's a raised hand function we're not going to be monitoring so if you raise your hand, I'm not going to be able to notice it. So again, please, Q&A. All right. So hopefully this visual will help you as if we were in person. This just makes me laugh so I have to share it. So this is our little group and we're really, really thrilled to bring you this content. So now I'm going to turn it over to Susan Xiao who's going to actually give you a little intro to a little bit supplemental information, I would say, to interpretation of variants and then we'll go into some cases from Elaine and Mark. So thank you very much. Susan? All right. Great. Thanks for that introduction, Janelle. So before we go right into the cases, we thought it would be helpful to have a little bit of a primer on how to interpret and report variants, both germline and somatic, from NGS assays. We didn't specifically cover this in our prerecorded lectures but we thought this would be helpful before we actually jump right into the application and we see how we use these criteria and the evidence to classify the variants in the cases we're presenting. So there are a lot of resources that are very helpful for investigating variants. There's many databases, publicly available databases. Certainly the intranet is a very helpful resource. You know, sometimes just Googling a variant, sometimes actually there's patient support groups or information on Facebook even that can be sometimes helpful and you know, really helpful to find other people with a similar phenotype. In silico algorithms also provide helpful evidence and then other resources that can be extremely helpful with some very rare genes and variants. Other laboratories that may have cases or patients with variants in the similar genes. Journal articles and of course experts. All this information, it can be very helpful to help us evaluate a variant and determine if it's pathogenic or benign. Oh boy. So I just wanted to highlight a few of these resources. So one of these resources is NOMAD, which is a population database. This has over 125,000 exomes and 15,000 exomes. And 15,000 genomes. So very good information about allele frequencies and cell population allele frequencies. And you can also has quality metrics and IGV views. So very useful resource frequently used. Another fantastic resource is from ClinGen and ClinVar. So ClinGen was launched in 2013, supported by the NIH. And this is a centralized resource that defines the clinical relevance of genomic variants to really build this genomic knowledge base. And so there's several projects within that. And I think it's still in development, but a very excellent resource for expert knowledge about genes. There's also ClinVar, which is a resource that where people can submit their interpretations and the evidence that they use to classify a variant. And you can see how different labs have classified it, if there's a consensus or not, with links out to different databases. So these are some really commonly used databases. In terms of in silico prediction algorithms, again, these also can provide very helpful information. So we have those that help with missense prediction. And these often use information such as evolutionary conservation, protein structure and function to predict the effect of the change. There are splice site predictors that use neural networks or Markov models to predict the functional, to predict whether a change in a splice site will affect splicing. And then others that use nucleotide conservation prediction. So how do we take all this information from all these different databases and resources and other laboratories and actually use it and synthesize it to classify variants and to try to do this in a reproducible way so that how one laboratory classifies a variant is similar to how another laboratory does? Well, there are these fantastic guidelines, standards and guidelines, which is a joint consensus recommendation of ACMG and AMP. And the reference is shown here. And basically what this does is it takes all these different lines of evidence, some of that which we've talked about. So population data, the in silico predictions, functional data, if there is any from the literature or from a laboratory, how the variant segregates with disease, if the variant is de novo or not, and information about the allele, if it's in trans or in cis, and all these other databases. And basically taking these different lines of evidence and what information they provide with you, use that to help guide you on whether a variant is benign or pathogenic. And this is showing the scoring system to classify a variant where you use the different lines of evidence you have and how strong these lines of evidence they are to put a variant to one of these five categories of pathogenic, likely pathogenic, benign, likely benign, or uncertain significance. So those were the guidelines for classifying germline variants. For classifying somatic variants, it's a similar idea, but it is a little bit different. And I find when I'm moving from a germline case to a somatic case, I do kind of have to shift gears a little bit. The basic principles of using evidence to help classify a variant are the same, and often a lot of the same databases are used. But, you know, whereas the classification is a little bit different, you know, for our pathogenic variants, we're trying to categorize for somatic testing into, you know, does this help us with diagnosis, prognosis, or is this predictive of a therapy? And there's also slightly different lines of evidence that we use in terms of clinical trials and, you know, case reports, preclinical studies, and, you know, guidelines from a different set of organizations, such as the FDA or NCCN or other, you know, professional society guidelines. Sorry. And so this is how somatic variants are classified. And so these are AMP-ASCO-CAP guidelines. The reference is shown below. Basically, the somatic variants are put into one of four tiers. Tier one is variants of strong clinical significance, where there's strong evidence that, you know, this variant is therapeutic, prognostic, or diagnostic. Tier two is variants of potential clinical significance, where there's some evidence, but a little bit less. Tier three is variants of unknown clinical significance, and tier four is benign or likely benign, and these are generally not reported. So I think that's it for the little primer on how to classify variants, and I'll turn it over to Elaine for case one. Thanks a lot, Susan. I'm going to go ahead and just start presenting this case, and what I'll try and do in the context of this is to tie together some of that primary information that Susan provided to us in the introductory remarks. So here's just information on this patient. She came to our attention as a two-year-old female with a family history of schizophrenia, epilepsy, and mental problems. Physical descriptors about the patient are listed bushy eyebrows, coarse face, thick, everted lips, and broad nasal tips. And the top three clinical features that were provided to us by the medical geneticist who recommended her for testing were global developmental delay, hypotonia, and failure to thrive. As is common as a precursor to exome sequencing testing, which is what I'll present results from in just a moment, there was some additional testing shown here that was performed initially, and the chromosomal microarray as well as the Fragile X SMA in vitae panel, including mitochondrial DNA, all were negative or non-diagnostic, and so this patient then proceeded to our clinical exome testing. What we were able to identify in this patient is shown here in IGV view, the proband data on the top in IGV compared to mom and dad down below. This is a 43 base pair deletion, so pretty significant size deletion in this gene, ARID1B, and this is causing a premature stop of translation, occurring about 37% of the way through the full-length protein within exon eight. This gene has 20 exons in total, and as you can easily see from the IGV view here, this is highly unique to the proband, so a de novo alteration. This slide is not projecting correctly, but what I wanted to do was to get you to take a little bit closer look at the insertion deletion event here, and this is really in the context of primarily focusing on how difficult and important it is to be able to detect indels in NGS data, so one thing to note here is illustrated is that given the blue and red strands, we want to note that the deletion is being represented in both redirections or on both strands of the DNA, and indeed that's a high-quality metric here. I also want to just point your attention to this very important panel along the top, which is showing that in the area of the indel, there's a decrease in coverage of the mapped reads. These are 150 base pair paired-end reads from Illumina sequencing, and we commonly see this in areas of insertion deletion just because of the randomness of fragmentation going into the next-gen library, as well as the fact that, in principle, a deletion that's this large, even with 150 base pair reads, is still a little difficult to map accurately, and so we end up discarding a lot of the reads that would normally be increasing the amount of coverage that you're seeing in that particular area that's represented, and so the take-home messages here are that the analytical pipelines in the clinical setting must have algorithms that can successfully map reads to indel variants and identify them. Often, these particular analytical approaches will detect the possible mapping of an indel region and then will go back and essentially fine-tune that mapping so that the accuracy is as high as possible. The other thing to point out in this particular case is that we have to have the adequate read coverage depth to maximize the detection of all of the variants that we would like to detect in these clinical assays. Okay, so next, I want to reflect again on some of the important material that Susan just covered, so we want to tally up the variant criteria in this particular indel that we've identified, and so here we have a combination of computational predicted data telling us this is a PBS1, very strong, likely pathogenic variant. We have confirmed the de novo inheritance is PS2, very strong based on the comparison to mom and dad from the same assay, and we have also looked carefully at the population database that Susan mentioned, NOMAD, to identify that this is largely, if not exclusively, absent in the population database as a PM2 moderate. Okay, so it's time for me to hand over to Janelle, who's going to provide the polling question. All right, so based on the applied criteria of PVS1, PS2, and PM2, what is the best variant classification? Okay, we're at about 60% of people voting, so I'm going to end the polling, and I'm going to share the results. So what do you think, Elaine? Well, I think that about 65% of people got this right. This is strong evidence for a pathogenic alteration based on all of the criteria and that summation of critique that Susan presented from the ACMG AMP guidelines. So this is just now a summary of what we just went through in terms of the different levels of classification. And indeed, this is, as I mentioned, a pathogenic variant. What you can see illustrated here is just the gene ARID1A, or B, rather, with all of the different types of alterations, which, as you can see, really predominate in nonsense and frameshift alterations, which would truncate the protein prematurely and probably make it very likely to undergo nonsense-mediated decay. So just a little bit of information here about ARID1B. This is located on chromosome 6q arm. It's part of the SWI-SNF family of protein complexes that regulate gene expression, influencing chromatin remodeling. And these complexes, as many of you know, are involved in many processes, DNA replication and repair, control of cell growth, division, and differentiation. It's shown here is how the ARID1B subunit binds to DNA directly and helps to target the SWI-SNF complexes to the chromatin location that is in need of remodeling. And you can see the reference that we used for these colorful diagrams in information shown below. So just another piece of how one would reach out to the peer-reviewed literature for information on the potential impact of this alteration that's been identified. The diagnosis, then, in this case, upon further evaluation of databases, is Koffin-Cyrus syndrome. As you can see here, there's a lot of locus heterogeneity in CSS. There are a variety of different pathogenic variants that have been identified across multiple genes, including genes that are involved in these different aspects of DNA remodeling. In ARID1B alone, about 69 pathogenic variants have been identified. So in addition to locus heterogeneity, there's allelic heterogeneity. And as you can imagine, as I'll show in just a minute, quite a variability in phenotypes as well. As we've already talked about, this is an autosomal dominant mode of inheritance. And most commonly, these are identified as de novo alterations with aspects that include learning disability, speech impairment, developmental delays, much like we heard in the primary case presentation for this child. There are also some congenital anomalies, cardiac, gastrointestinal, and nervous system. There are some congenital anomalies with degeneration of the corpus callosum, feeding difficulties, eye and hearing abnormalities, and some hypoplastic or absent pinky toenails and fingernails, as well as distinctive coarse facial features for these kids. So this slide just shows representatives of some of the different genes that are involved in Koffin-Cyrus syndrome. As you can see from left to right, the faces as well as these particular fingernail and toenail hypoplasias are evident. And I think it's quite interesting to see the variability, especially in the faces, around different genes and their manifestation. And then in particular, here are a group of kids just with Koffin-Cyrus associated with ARID1B pathogenic variants. And these phenotypes are very closely related to the ones that I already named out for this particular child. So just to finish up here, the Koffin-Cyrus syndrome management involves two things. One of these manifestations, and you can see that there are a variety of different treatment approaches depending upon the seriousness and variability of the phenotype. And then indeed, there's also, shown on the right, some aspects of surveillance as well, including evaluation yearly to assess developmental progress, GI involvement and evaluation to make sure that feeding and weight gain are appropriate, and then follow-up for ophthalmologic or audiologic abnormalities as well if these are involved. So I'll just finish this particular case, and I want to acknowledge the folks at Nationwide Children's who put this information together for me, including Marilena Mellis, who is one of our clinical fellows, Christy Lee and Marian Matthew, who are clinical directors here, and Ashita Degwala, who's one of our excellent variant scientists. All right. Thank you, Lynn. Time for questions. So there's not a specific question to this syndrome, but here's a question related to, you know, obviously someone that would have this syndrome comes in likely for multiple congenital anomalies, right? And you saw this indel, which is at the level of sequencing, but how do you work up a patient that has multiple congenital anomalies and you have a high suspicion for genetic disease? What are your thoughts on the utility of completing chromosomal microarray prior to doing whole exome sequencing versus doing the CNV detection using whole exome sequencing? This is a question from Amy Breiman. I think it depends to some extent on the assay. We know that we can evaluate copy number alterations, in particular from exome data. I would argue that, you know, the location of known human genes, which of course defines the exome, is never as regular, for example, in terms of their placement along the chromosomes as what you might imagine probes on a SNP array that would be used for microarray-based CNV detection might be. And so the level of resolution might be compromised for exome alone, and therefore the microarray might be, you know, an appropriate place to start. There are obviously challenges with microarrays as well. One of the ways that we have actually approached this is that in addition to having exome-specific probes in the capture reagent, we have the ability to add in chromosomal tiling probes as well. We often do this in our cancer assay for children because, of course, chromosome arms losses and focal gains and losses as well can be both diagnostic and prognostic, and so we want to capture that information when possible. And so that gives a much higher resolution along the chromosome arms so that you would be able to compensate for not doing a microarray if you were doing the exome capture with the augmented probes in there for a copy number alteration. That's one way to answer the question, but I think those are kind of our current thoughts around why not do both. And you know, would one assay with one cost and one turnaround time, in essence, you know, be the right way to go? I don't know that we and others are quite there yet, but I think it's definitely a direction that we're going in. Real quick, one more question for this. How did you make the determination of the reduced protein in this first case? This is from Gail Vance. I would say that we did not make a determination of reduced protein. One could certainly try to look at RNA expression from the blood. We're often doing this in our rare genomics cases, but not in the clinical exome. You know, it's a bit of a mixed bag because not all genes are expressed in PBMCs, although this one obviously probably would be. But I think it's also a bit dodgy to try and pick up on nonsense-mediated decay, right? Because what you're looking for is something that ostensibly is not there. So I think here, really, it's more the severity of the alteration, the fact that these have been reported in the peer-reviewed literature, and a really nice overlap with the patient's phenotypic manifestations that really drove that diagnosis. I'm going to add one more since it's an easy question. Fang Li is asking, usually a DEL of 150 base pairs is difficult. What software was used in this case? So this is Mutec2, which is part of the GATP pipeline, and it does do that specific indel re-alignment functionality that I mentioned. All right. Okay, so the next case I'll present, Janelle named Finding Two Needles in One Haystack. I'm pretty sure you named this, Janelle. And this is, again, a whole exome sequencing clinical diagnosis of mental retardation defect with the second needle and unexpected atypical secondary finding. So hopefully you'll find this quite interesting. This is a 16-month-old male. You can see a number of different phenotypic manifestations for this child and the reasons that his case was brought to our attention, including microcephaly, inappropriate fusion of the skull, global delay, scoliosis, et cetera. So this was subjected to both microarray and whole exome trio clinical assay. Here's the results from the microarray analysis showing a chromosomal gain quite reasonably large within chromosome band of 5p13.2. It contains some known genes, but really nothing that we could associate with a defined clinical phenotype. And so this is a really clinical significance unknown. If we move then ahead to the findings from whole exome sequencing, the primary sequencing result is shown here. This is PPP2R1A, a chromosome 19 gene with a variety of different attributes here. It's de novo variant in the child. It's absent from large population databases like NOMAD, as we heard for the last variants. There is some in vitro functional analysis that indicates that this variant that we've identified, which is a simple point mutation, a non-synonymous mutation, has reduced protein activity. And then a fourth criterion, this is predicted to be damaging by use of multiple functional prediction tools, which Susan already talked about in her introductory remarks. And so the variant interpretation here is likely pathogenic. This is just additional information for the ACMG evidence that I just walked you through. Again, a de novo alteration, as you can see in the comparison of proban to mom and dad. And then ClinVar information listed down here with some information on either side that's been gleaned from the NOMAD database. In addition, if we look at the protein paint representation of this gene, as shown here, we can see that there are a couple of variants that have been identified in the general neighborhood of our variant, which is circled here in protein phosphatase 2. These have been identified previously as potential drivers of endometrial or ovarian cancer. And then this R13883, or other Q, is associated with MRD36 and has been seen in another patient that we also studied in our whole exome sequencing clinical assay. So this just shows, then, again, a summary of some of the clinical features of patients who have these de novo missing variants in this protein and are classified with a mental retardation 36 phenotype in diagnosis. So you can read through these here and see that there is some variability, again, depending upon the different variants that are in the mix for this same gene. We did have an unexpected secondary finding in a medically actionable gene, so I'll just show you that here. This is a gene CDL, again, a point mutation, but interestingly, at a lower frequency than we would anticipate normally for a de novo variant. So this is de novo, of course, but it's mosaic. And so there's a variety of information associated with this variant in CDL that have been identified. As I mentioned, de novo mosaic correspondingly upregulates critical genes in the RAS pathway and diminishes E3 ligase. It is a mutational hotspot, a conserved tyrosine residue that's in a critical linker region of the protein, and then finally, again, predicted to be damaging by multiple functional prediction tools, as we heard for the previous variant. So this is an interesting situation. You can see here the recount data for the proband in the CBL variant here, this tyrosine 371 histidine. And you can see that the C residues are present at about 11%, and the predominant T is the wild type. If we go look again to compile these data for the ACMG-AMP evidence, you can see information shown here about the location of these alterations. This is really small on my screen, so I'm going to put on my reading glasses here. Sort of grouped together here, the fact that from the database, this is interpreted as pathogenic. And there's some additional information down here from OncoKB, which is another clinical cancer database that we often look to in the consequence of cancer-related variants like CBL. And so this is just, again, some of the information that we put together to begin to try and understand the meaning of this variant that we had detected as a mosaic. There's additional evidence shown here from the peer-reviewed literature. I think the most important thing to point out here is that germline CBL mutations can cause both development abnormalities. Certainly, this matches with the child's phenotype, but they also predispose to JML. And so we want to be cognizant of this as we follow this child over time. And I guess Janelle's going to pause here, and we're going to do a polling question before I finish up. What would be the appropriate follow-up testing? The voting's a little slower, I think, because there's more words. We're at about 50%, so I'm going to end polling and share results. What do you think, Elaine? Yeah, so the correct answer is the one at the bottom, the buccal swab and focused PCR assay. And we really need to understand the aspects of this using a different tissue in terms of whether it's truly a de novo variant or is just unique to hematologic tissue. And so that was what we did next. Thanks to you who voted for number one, which was my decoy question. I just wanted to point that out. Okay, so what's additional information about CBL? Well, if you look in NOMAD, it's quite interesting, actually. As you can see here on the far left, really the only individual, so across populations, exceedingly rare. And really the only individuals we're seeing this in are individuals of, I would say, advanced age, since I fall into one of these classes. But this is highly unique to individuals of older than 50 years of age. The top is from exome data, the bottom from genome data. And the reason for this is because these are individuals who are exhibiting what's actually now a pretty well understood phenomenon, but was new a few years ago, and that's called chip clonal hematopoiesis of indeterminate potential. This is really a situation when a blood, a heme stem cell actually acquires a single alteration. It's not cancer, but it gives it a growth advantage. And so you see that preponderance of that variant present in the individual's blood, although they're in general healthy, but potentially predisposed to longer term challenges from a health perspective that could include the development of leukemias. This is just in the center, then, some of the CBL variants that have been identified in consequence of clonal hematopoiesis from the NOMAD database. And the point of the last question and the result from that is that we used a buccal swab DNA to show that this mosaic in this individual does seem to be confined to the bloods. This is not a constitutional defect in the de novo setting. It's truly in the heme compartment. And this is really where we want to sort of finish up this discussion, because it's kind of interesting to look, at least it was for us, recently at the frequency of detected mosaicism and the genes in which it occurs. And so what I'm showing here is some published data on the top, and from an NVTAE presentation at ACMG last year, talking about their clinical cohort of over 400,000 individuals tested across 245 genes, and the numbers that they were able to show in terms of mosaic medically actionable variants being about 1%. In our experience, this is our cohort of 357 patients studied across about a year's time, having about a 32% diagnostic yield overall, but of those individuals attested, about 4.3% actually had positive test results that were mosaic. The genes that were identified and the variant allele fractions are shown here, and in a couple of cases, the mosaic parent transmission to the child, including one CSS case. And so this, I guess, raises a little bit of a question about ACMG reporting, which of course, you know, is confined to a gene list that's shown here, but indeed, we have, as shown below, identified other genes with alterations in them that we would consider to be medically actionable variants, and so we're going to look at that. So, we're going to look at the mosaic variants that are identified, and we're going to look at the mosaic variants that are identified, and we're going to look at the mosaic variants that are identified, and we're going to look at the mosaic variants that are identified, and we're going to look at the mosaic variants that are identified, and we're going to look at the mosaic variants that are identified, and we're going to look at the mosaic variants that are identified, and we're going to look at the mosaic variants that are identified. This is Wes's typical diagnosis of an MRD36 with a pathogenic variant in PPP2R1A that is well known. Secondary unexpected finding was this de novo mosaic in the CBL gene seen only in the blood, not in the DNA from the buccal swabs. And the follow-up steps for this child will be surveillance every three months for JMML. So again, I just want to acknowledge the group at the IGM. Marilena, again, put these slides together. Three of our clinical directors, including our lead clinical director, Dr. Kathy Cottrell, contributed to this diagnosis and sign out. Amari Mori is one of our medical geneticists and, again, one of our outstanding IGM variant analysts, Vijay Jaraman. And also want to acknowledge our two genetic counselors, Sayaka Hashimoto and Maggie Stein. All right. Thank you, Elaine. One question that's come up a couple of times while you were talking, looks like Neng Chen as well as Ulrich Kappas are asking about how do you go about determining a mosaic status of a variant? Do you do a confirmatory assay? How do you know it's not just an artifact? So how do you deal with mosaic variants? Yeah. So, I mean, I think it's still addressed in the slide that I showed, which is really a careful quality control evaluation of the data from the whole exome sequencing, making sure that the coverage is reasonably high, that we have double-stranded coverage, and that the variant that's being detected, for example, is not at the end of a read, where, as I talked about in my introductory talk on NGS, you often see artifacts creeping in because of the increased error rate on those. I think the real challenge for somatic variants in terms of some orthogonal validation is what do you use? So typically, depending upon the percentage VAP, we could use Sanger, but ultimately Sanger falls apart when you have a really low-lying variant. And we're beginning to evaluate new approaches that don't rely on Sanger, but are NGS, but a different type of assay that could potentially be used to validate that. So basically, orthogonal testing is what we're looking for. Yeah, exactly. Okay. Yeah. Okay. Well, just based on time, what I'm going to do is go ahead and move on to the next case, but I do see there's more questions, so we'll answer some more at the end. Thank you very much, Elaine. Yeah, you bet. All right. Mark? Thank you. So this is a case that Janelle also named. She's quite good with these. This is a case for calling somatic surprise. Let me see if I can... All right, there we go. So this case is a 39-year-old man presents with acute myeloid leukemia. Here you can see in a peripheral blood smear, there are a lot of these immature blasts in circulation. And this got chopped off a little bit, but when we see a new patient with AML, one of the things we want to do is figure out what is the risk stratification. And the NCCN guidelines recommend a fairly comprehensive genetic and genomic evaluation of patients with de novo AML as a way to try to assess their particular prognosis. Here is their risk stratification. And as you can see, many of these are cytogenetic abnormalities that can either be detected by karyotype or FISH, but there are also several specific small nucleotide changes or mutations that can be detected by genetic analysis such as NGS. In this particular case, cytogenetics, both karyotype and FISH were normal. However, there were two distinct CEBP-alpha mutations that were identified. One was detected was a frameshift alteration with a VAF of 54%, and the other was an in-frame insertion with a VAF of 55%. Unfortunately, because this was done using short read sequencing, we are unable to determine whether these alterations are in cis or in trans. However, classically, CEBP-alpha mutations tend to be in trans. So you can see here the different applicon blocks of the sequence, and you can see that one alteration is on one group and the other is on the other, so we aren't able to phase them. Using some of the long read sequencing technologies Dr. Martis discussed in her recorded lecture, it would be possible to determine that these are either in cis or in trans. As Dr. Hsiao mentioned during the introduction, when we're looking at interpretation of variants in cancer, we often turn to the CAP-AMP-ASCO guidelines for interpretation of variants. And again, we see here there are four tiers of evidence. Usually we report the first three tiers, the strong clinical significance, the potential clinical significance, and the variance of unknown clinical significance. Because biallelic mutated CEBP-alpha is part of the NCCN guidelines, this meets criteria for being a Tier 1 alteration or of clinical significance in the setting of AML. Interestingly, as I said, most cases where you have two CEBP-alpha mutations, they tend to be in trans, but also they tend to have a fairly stereotyped pattern of mutation. There tends to be a cluster of N-terminal mutations, and these tend to be frame shifts or nonsense mutations, and those occur immediately following the first start site for the P42 isoform of this transcription factor. And then the second group of alterations tend to be in-frame insertions or deletions, occasionally missense mutations, and these occur towards the C-terminus of the protein, and these affect both the P30 and the P42 isoforms. However, because these early alterations tend to be frame shifts or nonsense, they tend to lead to a truncated protein which cannot be translated, and so we alter the balance so that there's more P30 in these patients. So here, again, you can see that when we have bilelic CEBP-alpha mutations, we shift the balance to increase the amount of relative P30 activity compared to P42 activity, and this alters the binding of CEBP-alpha to DNA and alters the transcriptional profile of these cells. CEBP-alpha is a known transcription factor that's involved in myeloid differentiation, and these changes lead to an impaired differentiation in immature myeloid cells. This patient was treated and responded very well to standard induction chemotherapy, but seven years later, he clinically relapsed. Again, he had a normal karyotype, but when he had his myeloid MGS panel repeated, he again was found to have two CEBP-alpha mutations. The first of these mutations at the N-terminus was the same as the prior, whereas the C-terminal mutation was de novo. So at this point, Janelle, we'll pop the polling question for the group. Okay. What do you do next? Okay. We're at about 50 percent. All right. Wow. It's a spread. What do you think, Mark? All right. So great. The majority did get the correct answer, although there was some variability here. So we talked about that in the guidelines, we talked about that in the guidelines, they talk about bioglialic mutations of CBP-alpha as being a favorable prognostic indicator. But there has been recognition of germline alterations in CBP-alpha as a predisposing factor to myeloid neoplasms, particularly AML. And in the setting of a hematologic neoplasm, it's really important if we're trying to get a source of germline that we use a tissue that will not have peripheral blood contamination. So testing peripheral blood should give the same results as we received on the bone marrow NGS panel. Buccal swabs, although they will contain a lot of the germline cells from within the mouth, have been shown to have a lot of infiltration by monocytes and other hematopoietic elements. And so using a buccal swab as your source of normal for patients with blood cancers can often lead to false positive results and misinterpretation of somatic variants as actually being or misinterpretation of a somatic variant and thinking it's germline. Because of this testing, the cultured skin fibroblasts is the best source of getting a true sense of germline in these patients. And because, as you'll see in some of the following slides, there is prognostic implications of having a germline alteration, it's important to recognize that this patient had the same alteration in two subsequent relapses of AML. And so not doing additional testing would fail to give this patient the best follow-up and care. So when these cultured fibroblasts were sequenced, they showed the same N-terminal deletion that we saw both in the initial and the relapse AML sample. This confirmed that this was in fact a germline alteration in CEBP-alpha. Okay. Okay. What's the diagnosis? Okay. We're about 50%. All right. Let me share. Great. So the vast majority of the group got the correct answer. And some of this is a little bit related to the classification in the WHO classification of AML and in which alterations take precedence over which other alterations. Generally, having a germline alteration takes precedence for the classification over the presence of the bioallelic mutation. So if we didn't know that this patient had a germline alteration, then the AML with bioallelic CEBP-alpha mutation would be the correct choice. And also if we had proven that the cultured fibroblasts were lacking that alteration, that would have been the correct answer. In terms of normal karyotype AML, this is in fact not a classification recognized in the WHO. It would fall under AML-NOS. And in the presence of a specific recurrent genetic alteration, you cannot make a diagnosis with AML-NOS because there's a more specific subtype in which you can land this. Why does it matter? One of the main reasons why this is particularly important is AML with germline CEBP-alpha mutation is highly penetrant. Here we can see several different pedigrees where there is an inherited CEBP-alpha mutation. And you can see that almost every patient or every proband or carrier of this alteration goes on to develop a new mutation. So, this is very important for family screening. Also, it's going to be important for donor selection if this patient is going to undergo a stem cell transplant. This slide may look familiar. It actually is very similar that there are common patterns of CEBP-alpha mutations in germline CEBP-alpha mutations as well. These tend to cluster in the N-terminus of the protein and abrogate the P42 isoform, whereas the acquired mutations tend to occur at the C-terminus of the protein. And that was the case in our patient. Also, as happened in the case of our patient, often patients who have a germline CEBP-alpha mutation at relapse will go on to develop a de novo C-terminal CEBP-alpha mutation. So, here you can see that there's a change from the original diagnostic and the relapsed sample in many of these cases, although not all of them, whereas the last case, it's identical. So, it's unclear if this represents a relapse of the initial clone or just a repeat mutation at the exact same position. And clinically, what happens with patients who have these germline CEBP-alpha mutations is that they have an alteration of CEBP-alpha within their hematopoietic stem cell pool. And all of these stem cells will have this alteration. And then a second hit of a second CEBP-alpha mutation leads to the outgrowth of this neoplastic AML clone. And once that clone is treated, these are very chemoresponsive. A second clone is able to develop a different CEBP-alpha mutation. And this process can recur multiple times, which is why it's important to select a donor who is shown to lack a germline CEBP-alpha mutation. And family screening is really critical if you're thinking about a related donor in these patients. Despite the fact that these patients have a very high risk of relapse, the prognosis for these patients is very good. Here we see patients who have spontaneous single CEBP-alpha mutations. This would fall under an AML-NOS group in the WHO. Those spontaneous double or the biallelic CEBP-alpha mutations. So these are the improved prognosis that we saw in those guidelines. And actually patients who have familial CEBP-alpha mutations have the best prognosis of all. However, interestingly, if you take the cohort of young patients under 40, you see that the spontaneous double and the familial cases overlap in their prognosis. So it seems that part of the improved prognosis is related to the fact that patients with these germline CEBP-alpha mutations presented very early age. But also, looking at additional studies of these young patients, we found that 7 to 11 percent of AML patients with biallelic CEBP-alpha mutations actually have a germline mutation. So it's important when you have a patient who presents at a young age, particularly under 40 years of age, to consider doing germline testing in patients with biallelic CEBP-alpha mutations to make sure that you don't miss this rare but important alteration. So just some highlights. Certain quote-unquote somatic alterations should prompt germline consideration because these impact the relapse risk. They could change who you consider for donor selection and can inform family screening. And for AML with CEBP-alpha mutations, if they have good prognosis, if it's a biallelic alteration, but it's important to recognize that there are two different subgroups within the WHO, and both of them have a similar pattern of alterations with the mutation clusters at the N and C terminus to alter the balance of p42 and p30 isoforms. I'd like to acknowledge Dr. Amy Treese for providing the details of these alterations in this case. She's the medical director of the Precision Diagnostics Laboratory at Children's Hospital Colorado. And now I'd like to open it up for questions. All right. Thanks so much, Mark. So I think there isn't a specific question that I'm seeing for this case, but I think what would be helpful is one person's kind of asking, you know, you guys have shown different evidence and, you know, said this is our conclusion, but one, Sadaf Sabah was asking about, you know, the final decision about pathogenicity, and this could apply to constitutional cancer, you know, pathogenicity of a variant and accumulating evidence. Is this done manually, or are there computational algorithms that sum up all the scores for you so that you can make and then just tell you the decision, or how do you do this? So in my laboratory, it's done entirely manually by trainees and then reviewed by attending pathologists. Other laboratories have variant annotation scientists who do this. There are variety of computational programs that are able to take this information into account. But as far as I know, every lab does do some kind of manual review for new alterations. I'd like to hear what others think about how this is done in their laboratories, because I think there is some varied practice, particularly in labs that have more sophisticated informatics capabilities. Sure. Elaine or Susan, did you have any, did you want to add anything? I would just add that largely this is a combination of databases that we curate in terms of variants that we've seen in the past and information coming in from other databases that we reach out to, many like what Susan presented initially. Most of this legwork is done by our variant scientists and then reviewed by our clinical directors. So it would be nice if it could become more automated over time. I'm sure that there are some commercial softwares that are aiming for that, but I think the majority of work is done by our variant scientists at this point in time. Okay. We do have one question here. In somatic testing, Mark, how are you incorporating the results of cytogenetic testing complex abnormal karyotypes into your NGS results? That is a great question. I think one area where it can be informative in particular is if you're thinking about cases that have P53 alterations, perhaps in lymphomas, you may see that by fish or by array, you can see deletion of 17P, which is the locus where P53 resides. And then you may find a very, very high variant allele fraction alteration in P53 if you have a mutation in the other allele. And so there's loss of heterozygosity. So really, this patient may have no functional P53 and in some settings, this may portend a particularly bad prognosis. I think it's important to look at the whole picture and the specific context of the particular cancer that you're looking at in somatic testing to understand which alterations impact the staging for prognosis, prediction of therapeutic response, and for diagnosis. It looks like a couple of you has a similar question. Renu Baha and Anup Tilak were asking about what's the BAF for the noose EBPA mutation and would a germline mutation BAF not be higher? So I don't recall the BAF for this one, and that's why it wasn't included in the case. But generally, when you are sequencing an AML bone marrow, you tend to have a very high blast count. And therefore, if that de novo clone has really taken over the bone marrow compartment and is all those neoplastic blast cells, then you might expect to see that you would have approximately 50% BAF for both your germline alteration and your de novo somatic alteration. If, however, you have a lower blast percentage, say 30% blast, you would expect to see that the germline alteration would have a BAF of 50%, whereas the de novo somatic would have a BAF of about 15%. But it varies a little bit based on the tumor cellularity in that setting. Makes sense. So one more question for this. Abha Panagana and Harold Bass are asking about, you know, donors. Is donor testing a common practice for such cases, or do you routinely check donors for CEBPA mutations? I'm assuming, you know, you could expand this out to other germline predisposition to cancer mutations. Is any of this kind of screening done? This screening is really only done in the setting where the patient is preferring to go with a related donor to make sure that they don't share this familial alteration. So if you're to make sure that they don't share this familial alteration, it generally is not done with routine screening of hematopoietic stem cell products. Many of our patients get just cord blood units or National Donor Registry units. Those do not undergo routine genomic analysis as far as I know, although we could foresee that happening in the future as the cost of assessing continues to drop. Yeah, I mean, you certainly see that, though, where you'll end up with donor, you know, cell leukemia, right? You're looking at the replacement of the bone marrow of the patient, and they have a whole different other, you know, mutations going on. You're like, what happened? Yeah, we've seen it. All right, that was great. Let's go ahead and move on to the last case. Thanks, Mark. Okay, thank you. So this next case is trying to look at one of the unsung heroes that we have as one of our metrics when we're looking at our NGS cases and trying to figure out what alterations are present. Dr. Martis already covered a little bit about how coverage can be important, but I'm going to hopefully convince you that it can be very important in somatic testing. So come on, there we go. So here we have a 44-year-old woman with no prior smoking history, has a lung nodule, and biopsy shows a lung adenocarcinoma. Here we just have an image of histology on the right. Again, NCCN guidelines are really the standard for profiling of new diagnosis cancer patients, and there are particular alterations that are recommended as routine testing, including molecular analysis of EGFR for mutation testing, as well as BRAF, and then testing for rearrangements of ALK and ROS1. In addition, the NCCN recognizes that there are some emerging biomarkers, such as MET amplification or events that lead to skipping of MET exon 14. This leads to constitutive activation of the MET proto-oncogene. Also, RET rearrangements and ERBB2 mutations are our emerging biomarkers, and tumor mutational burden, which we will not be discussing today, is another biomarker that can be used to predict response. This patient underwent fairly widespread testing by an NGS panel for mutation analysis. They did not have any specific mutations identified, and they had fish for ALK, ROS1, and RET rearrangements, and were normal in all of these. However, when we looked at the coverage pattern for the MET region of this patient, we noticed that there was something a little bit funny. So, on the left, we see the coverage of most of our patients where this is amplicon-based testing, where all the reads are lined up on one edge. Just based on sorting, we aren't able to see the next set of reads, which would be down here, would be below the area where we see the genomic layout of the MET gene. But there's a similar depth of coverage, both in the area where the reads are shown and the white space to the right where the reads are not visible. You see in our patient here, we have markedly reduced coverage in the area where we can visualize the reads, compared to the area where the reads are just sorted out of the frame of the image. And this is somewhat unusual. And if you notice that this is happening pretty adjacent to an intron-exon boundary. So, just reviewing how IGB shows the gene layout, at the bottom of the panes, you can see the large blue rectangle with the forward arrows. That is the exon. And immediately following that is intronic sequence. All right, so why is there reduced coverage in half of this region? Okay, we're at about 50 percent. So, let me share. Okay, it looks like people are pretty split on whether this is a local copy number alteration or this represents allele dropout. We'll talk a little bit about what allele dropout means in the coming slides. Both of these could be correct answers. I think that allele dropout is the better answer because this is a MET exon 14 splice junction mutational hotspot. And in this area, we often see very large deletion mutations, as you'll see in the coming slides. And therefore, I think that one of these alterations may have impacted primer binding, leading to dropout of the mutant allele and decreased coverage. This could represent a local copy number alteration, and this could be a deletion of one region. However, the most common types of mutations in MET and adenocarcinoma of the lung at this area are really related to those alterations that lead to splicing variants. And so, I think it's more likely to be an allele dropout rather than a copy number alteration. Low tumor cellularity doesn't seem to be at play in this case because we can very clearly see a large difference in coverage. And so, it seems that we have adequate tumor cellularity here to be able to detect the alteration that we're looking for. So, as I said before, low coverage in amplicon-based assays may reflect impaired primer binding. So, here we can see where the primers are set up for the different amplicons that are shown here. Here we see the amplicon where we have the lower coverage, and then here we have the forward and reverse primers for the amplicon where the coverage is higher, but the actual pileups are not seen. If you have a mutation here underneath the primer binding site, you'll prevent the binding of this primer, and it'll lead to reduced coverage in this region, which will give findings like what we saw, but you won't be able to actually amplify or sequence any of the mutant alleles. And so, it will lead to a wild-type sequence when you look for mutation calls. Comparing this to hybrid capture sequencing, because amplicon sequencing uses these PCR primers that are at set positions, it has a much higher risk of losing the ability to trap variants because you can have an alteration underneath this primer binding site, leading to failed amplification. In contrast, hybrid capture sequencing, because it takes randomly sheared DNA and uses hybrid capture probes, it's possible that you may have a slight decrease in coverage in areas where there's deletion mutations due to impaired binding. However, overall, you have a more representative coverage of the region, irrespective of the presence of different alterations. Okay, last polling question. What testing should be performed to confirm the diagnosis? Slowly building. Okay, we've hit about 50%. Here are the results. Great. So, the majority of the group got this right. As I said, this is a hotspot area for met exon-14 skipping events. And using RNA-based testing, you can be able to see the presence of the variant transcript where exon-13 is directly in front of exon-15 with the removal of exon-14 from the splice variant. So, RNA-based testing will confirm this alteration. Microarray will allow us to detect if there was a copy number alteration. However, in this particular instance, we have a relatively small alteration. Most of the amplicons that we sequence in amplicon-based sequencing are a few hundred base pairs, whereas microarrays really are only sensitive for alterations of about 25 to 50 kilobases, so much larger copy number alterations. And because, as we said earlier, we think that the tumor cellularity is adequate, repeat microdissection to enrich wouldn't provide additional information. Okay. So, here we have some of the sequencing information from an RNA-based fusion analysis. This is performed using anchored multiplex PCR. But you can see here that you have many reads where you have met exon-13 here is just supposed to met exon-15 here, and they read right through. And overall, looking at the sort of graphical user interface provided by the vendor software here, you can see that you have this oncogenic isoform of met where you have exon-14 is skipped in this product. And as I mentioned, there are a variety of different types of met exon-14 skipping variants, and this is from a review in 2015, and you can see that some of these are very, very large alterations, and they tend to cluster at both ends of the exon-14 and incorporate the canonical splice sites. So, the plus, minus one, and two base pairs are impacted by these variants, and some of them are extremely long. Our variant would have been here at the C-terminus of exon-14, and that's what we believe happened in this case, but we did not have another method that we were able to confirm this alteration at a DNA level. So, as I mentioned, exon-14 skip events in met are oncogenic, and exon-14 here is the site where SIBO can ubiquitinate the met protein and lead to its degradation. But by skipping exon-14, the ubiquitination site is removed, and SIBO is no longer able to ubiquitinate and degrade the protein, so you end up with a relative increase in the amount of the met oncogene present. In terms of the clinical response, chrysotinib has been shown to have activity in med-exon-14 altered lung cancers, and here you see that with this waterfall plot that most of the patients who received chrysotinib in this study had at least a partial response when they received chrysotinib therapy. And because of this, both its inclusion in professional guidelines as an emerging biomarker and well-powered studies with consensus on how this works, biologically, we're able to classify this as a Tier 1 variant as well for the met-exon-14 skip event in non-small cell lung cancer. So, the highlights. Untiled PCR-based libraries can suffer from an allele dropout. One way that you can correct for this is if you have overlapping primer sets that create what we call tiling. And also, splice events require specific design. So, many of the splice sites and the splice site alterations impact intronic sequences, which are not often covered in some of the standard-targeted NGS assays and may not be well-covered in exon sequencing. And because of this, if you're interested in looking for splice variants, you may need to extend your coverage into the intron when you design the assay, or you may need to consider using RNA-based testing as an adjunct to help confirm cases that either lack an alteration by DNA-based testing or have strange findings, such as in this case. And then for the clinical significance, met-exon-14 skip mutations in lung cancer lead to activation of MET and also lead to response to presotinib therapy. I'd like to thank Kurt Davies, who's one of our development scientists in the Komoko Laboratory, for providing this case, and see if anyone has any questions. All right. Thank you, Mark. That was really great. We're a little bit more short on time, so I think instead of targeting it just to this case, we're going to move into the broad question-and-answer portion. So I'm going to stop sharing my screen so you can see all the panelists. Hello? All right. So one question that's popped up a few times is really the whole concept of Sanger, right? And so what are the differences in Sanger versus NGS methods and considering the ability to detect a point mutation on Endel and why do labs still do Sanger confirmation routinely? And I was going to aim this question at you, Elaine. Thanks, Janelle. I mean, I think it's an important question, and we've had, as a group, a lot of discussion around it. And you've seen some examples in the data that we presented even for these case studies. So in cases that both Mark and I presented for Somatic as well as Germline, you know, these large Endels, we all know how they look on Sanger traces. You know, the sequence reads up to a point at which the Endel starts, and then you see the reflection of a garbled sequence that's essentially representing all of the products that were in that pool of products that were sequenced and loaded onto the gel. And so I think in this case, you know, I don't personally understand a compelling reason to confirm an Endel variant using a Sanger approach because who can interpret that data other than to say, yeah, this is what an Endel looks like? You know, not to impugn Sanger, but this is just a place where Sanger clearly falls apart. I also mentioned the challenges even with simple point mutations in the context of these mosaicisms that I pointed out in the second case. But of course, as I'm sure Mark could confirm as well, in cancers, when we really are looking for very low variant allele fraction representatives such as a known resistance mutation, for example, in a lung cancer patient who's going to go on EGFR therapy, but, you know, likely come up with some sort of a point mutation to become resistant to that therapy, T790M, for example. You know, these are going to be very, very difficult to detect at low levels, which is where you really would like to detect them because in the current scenario for lung cancer, then there are a variety of other targeted therapies that you might go to, such as third-line therapies that were actually designed to work around the T790M. So again, I think in the Sanger setting, you're really going to be in a delicate situation of not having the sensitivity to detect something that's at that low prevalence. That's true. I mean, I think there's a lot of assays that have VAFs detecting down to 5% for these somatic mutations, and Sanger just can't get down below 15-ish percent. So it can't even confirm a lot of those mutations. So then I think it's really important to, you know, make sure that your filtering strategies are there so that it gives you the confidence in the NGS, and obviously you develop this as you validate a clinical test anyway, which I think Mark focused a lot in his recorded talk in terms of studying samples that you know from a variety of different sources to really make sure that, you know, you're filtering out artifacts as best possible. Nothing at the end of the day really replaces that visual examination of the data through IGB, if you will, but, you know, that limits the number of sites that you really end up having to look at carefully. So we're not going to be completely replaced by robots yet. We're not. Okay. Hey, Susan, how about this? How do you differentiate false positive calls from low-percentage mutations in tumors? That's a great question, and I think that kind of follows some of the things that Elaine was talking about in terms of, you know, how do we have confidence in calls that have low variant allele attraction? And I think, you know, part of the answer is going to be technical in terms of the validation that was performed and, you know, how low a variant allele attraction were you able to reliably detect. But there are also other considerations to think about. You know, with variant calling, you're really talking about signal-to-noise, right? So some approaches you can use to kind of help you in this, one is going to be coverage depth. So if you sequence to a higher depth, I mean, to a limit, but if you sequence to a higher depth, you know, that gives you a little bit more signal rather than noise. Another approach that some people like to take is to use multiple variant callers, and you can either, you can take the consensus call, and that can kind of help reduce some false positives. Other approaches, if you really need very low sensitivity, you can consider using UMIs, or unique molecular identifiers, and this will allow you to do error correction, which, again, will help, you know, increase your confidence in a call. Normal tumor sequencing is another way that can help because you can sort of use the normal as a correction. And finally, just in terms of any variant, you always just want to be aware of some problem areas in terms of sequencing, so repetitive areas, or if they're pseudogenes, things like that should definitely raise your suspicion about a call in one of these areas, so you should check the mappability of a particular region, and it may be confirmed by an orthogonal method, be that Sanger or some other more sensitive method. I'm a dreaded Sanger. You can't get away from it. Okay, I'm going to direct one to Mark. When do you need to revalidate an NGS platform or assay, and does this play into your strategy to how you set up your initial assay, such as, like, targeted panel versus whole exome? So, the quick answer to part two is yes, it definitely impacts the assay design, but generally, I think of the validation as any time you make a significant change to the assay, you need to revalidate it, so if this means you want to alter the content of your assay, then you need to revalidate your assay by either adding additional capture probes or different PCR primers if you're using an amplicon-based approach. If you're making major changes to the bioinformatics, you also need to revalidate. However, this validation can be done as a quote-unquote in silico validation, so rather than going back through and starting back and repeating all of the wet bench work that was done initially, it's possible to take the raw data from your sequencing instrument and run it through both the initial pipeline and the new pipeline and compare the results. Generally, when this is done, it's to resolve any type of problem with the initial pipeline, and so it's important then to do targeted sequencing of any discrepancies that are found between the old and new pipeline to show that the new pipeline has improved the performance of the calling and not sacrificed any sensitivity. Because of this, this really does impact panel design, and so some of the larger labs are going with whole exome sequencing as a backbone for many of their different tests, and then they're picking out specific genes to use as targeted panels for particular indications, and the depth of sequencing here is going to be influenced a little bit by what your goal is. So if you start with a constitutional exome, you may carve out a lower depth panel that you then sequence, and you make subpanels for specific indications for constitutional disease, whereas if you're doing a cancer exome, you're going to have to sequence at higher depth, and it starts to get a lot more expensive. So you may want to go with a sort of mini exome of 500 to 1,000 genes, where you sort of take the most clinically actionable genes and then carve out subpanels from that when you're thinking about the assay design. All right. I'm going to try to fit in a few more questions because there's so many good ones. What level of whole genome coverage is likely necessary to generate chromosome maps similar to those used by microarray analysis? This is from Justin Schlead. I think I can try to answer that. I mean, typically, as I mentioned, our approach is to add in extra probes that just allow a tiling through the chromosomes. It doesn't add much space. Gives you about 30 kb resolution on average for the reagent that we use, which is available from IDT, and to which I have no commercial ties, just to be clear. But if you really want to go for the whole genome coverage, that can be actually a good approach because of the evenness, especially if you can get away with using a low or no PCR-type library preparation method. And typically there, I think most people would try for somewhere in the neighborhood of two- to five-fold coverage to get very high-resolution copy number alterations. If you wanted to get structural variants out of whole genome, you'll obviously have to go to higher levels to achieve that, depending upon tumor cellularity, if it's for cancer and other things. So, but for microarray, you know, for copy number-based LOH, et cetera, you could probably get away with two to 5% in specialized algorithms. Kind of an interesting question. How are exon numbers determined? Often there are multiple transcripts and exons that are assigned differently depending on the transcript. So this is a tough one. I know, I wasn't sure if I should ask it. Well, yeah, I mean, it's super important, right? Because there are multiple transcripts. We typically take the most abundant one and there's a new sort of designation called the main, M-A-N-E, which provides the majority transcript that's been identified from sort of evaluating, you know, transcription in multiple large databases. And we typically rely on that main designated transcript as the transcript to annotate. But of course, as a big fan of sequencing DNA and RNA, you know, you can learn a lot from an individual patient if you had the time, in terms of what their, you know, transcripts look like. And we talked about why that can be important in a variety of different settings. So usually there we're looking at something very focused as opposed to the whole transcriptome. If that makes sense. Just to be clear. Okay, again, there's still some great questions, but we are at time. So I'm gonna have to wrap up. So a couple of real quick business things. To claim your CME credits, please complete your survey online on the ACMG digital site. And this webinar is being recorded and will be posted. We can definitely put the slides on as well. I don't think anyone has any problems with that. So no worries, we'll get those to you too. Importantly, we have one more webinar as part of the ACMG 2020 meeting coming up on July 20th from 3.30 to 5 p.m. Eastern. And this is on neurofibromatosis, new frontiers in NF1, NF2, and schwannomatosis. So I just, I really wanna thank everyone for attending today. And this has really been a wonderful experience for me, working with these great people. I can't say enough. We've had a lot of fun, really, truly. So thank you so much to the panelists and thank you to everyone for attending. Bye.

2-year-old female

family history

whole exome sequencing

ARID1B gene

Coffin-Cyrus syndrome

developmental delays

specific abnormalities

16-month-old male

microcephaly

PPP2R1A gene

mental retardation defect 36 phenotype

mosaic variant

NGS