I'm really excited to share with everybody today the cutting edge enhancements we've made to our exome platform that enable us to provide affordable, comprehensive exome sequencing to all patients and how these enhancements allow us to launch a first-of-its-kind case-level exome reanalysis program that is included in the cost of the test. So what is exome sequencing? As opposed to panels that target a set of genes related to a condition or phenotype in question, exome sequencing simultaneously interrogates the vast majority of all coding exons. This ends up being about 250,000 exons or 1.5% of the genome. By identifying changes in the DNA across the exome, we can offer a comprehensive analysis to pinpoint an individual's genetic diagnosis. There are several situations when exome sequencing might be considered for a patient. The first is when their condition is associated with many genes, and this can be the case for situations involving intellectual disability and neurodevelopmental conditions. The second example is when a patient is a product of a consanguinous unit, and in this situation, exome sequencing allows us to simultaneously interrogate many genes associated with autosomal recessive disorders. Exome sequencing may also be considered for situations where a patient has multiple diagnoses or if there's involvement of multiple unrelated organ systems. And again, in this situation, the list of differentials can be rather large. Finally, exome sequencing is often recommended for patients who have already undergone a diagnostic odyssey without a definitive diagnosis. So perhaps previous genetic testing has been done before or been negative or inconclusive or other diagnostic assays has been unable to identify a diagnosis. Many studies have now been published in the literature demonstrating the impressive yield of exome sequencing as high as 20 to 40 percent in patients with neurodevelopmental disorders, multiple congenital anomalies, and syndromic clinical presentations. Although exome sequencing has been instrumental in the diagnosis of individuals with rare genetic disorders, there are several challenges that stand in the way of its full implementation in clinical care. The first is the constant evolution of gene phenotype information. Almost every day, novel disease genes are published in the literature, and the phenotypes that are associated with known disorders is also constantly expanding. Because of these issues, it's quite difficult to perform a reproducible and unbiased evaluation of genetic variation in the context of the proband's phenotype. The interpretation of exome data requires manual review and judgment of variance in gene-disease relationships, and bias does exist due to the level of experience and knowledge that the individual performing the analysis has on various genes and disorders. We also know that there's a need for consistent evidence-based gene-disease curation and variant interpretation frameworks. In order to perform a complete analysis, we need to have up-to-date gene, disease, and phenotype curation information, and also a semi-quantitative variant interpretation framework that is consistent with the American College of Medical Genetics and the Association for Molecular Pathology Guidelines. While the guidelines that exist today are an excellent framework, it is our experience that refinement is required for clinically meaningful variant interpretations. Finally, one of the most difficult issues to address in the age of clinical exome sequencing has been developing a reanalysis framework that allows laboratories to routinely apply learnings from newly published information and internal data to help solve previously negative or inconclusive exome cases, and more importantly, to do it at scale. So, in order to address the challenges of exome sequencing for rare genetic conditions, InVitae's clinical exome analysis combines automated gene and variant curation with expert evaluation for fast, reproducible, and accurate exome interpretation. This slide outlines the different components of InVitae's exome sequencing test, and in the subsequent slides, I'm going to share more detailed information on how to I'm going to share more detailed information about our state-of-the-art interpretation process. Shown at the top of the diagram in teal is the wealth of information that gets fed into InVitae's moon. Moon, which is shown in blue, is InVitae's intelligent engine that integrates and weighs genomic and phenotypic data. At the bottom of the figure in green is an outline of our reporting process where our expert clinical staff ensure accurate, clinically relevant reporting for each and every patient. Clinical information is important for a phenotype-driven test such as exome sequencing. So, in the next couple of slides, I'd like to review with you InVitae's clinical data gathering and portal capabilities that help clinicians to ensure that a complete clinical picture is obtained for all patients undergoing exome sequencing at InVitae. Here's a screenshot of our ordering portal. Our exome can be ordered as a proband-only test, as a duo test with a proband and an affected parent, or as a trio. I do want to say here that trio testing has been shown at our lab and other laboratories to produce the highest yield. You'll see at the bottom of the slide here that each sequenced individual is given the opportunity to opt in or out of secondary findings analysis, which focuses on 59 genes associated with clinically actionable genetic disorders. We determined which variants to report based on the molecular strength of the variant and the clinical fit of the gene disease associations with the patient's phenotype. Therefore, it's important that we receive as much clinical information about the patient as possible. To that end, we provide multiple ways for providers to submit clinical information. Shown here is another screenshot of the portal. At the very top of the screenshot, you can see where free text can be entered for additional clinical information, and we also provide various mechanisms to upload previous results and family history. So, for example, you might be able to copy and paste a recent clinic note into the box, and we have natural language processing tools that systematically identify human phenotype ontology terms from your clinic note. We are also able to obtain pedigrees and family histories, upload previous genetic testing results, and also any previous diagnostic results that the patient has undergone, for example, an MRI, EEG, etc. We also have a number of online ordering resources for the clinician. The first shown here is our gene coverage tool. This allows clinicians to review any technical limitations for the genes of interest before ordering. For example, you might be able to review if a gene has low coverage or if there are any C and V gaps that are present before the gene, and this can all happen before ordering. Within the portal at the time of ordering, clinicians are also able to specify any genes of interest that might be on their differential. In addition, clinicians can also provide us with additional phenotypic information using our HPO selection tool. So here, clinicians can enter their own HPO terms. You can search this tool using features by organ or system, and I do want to note here that the clinic notes that are provided in the ordering portal at the time of ordering are also reviewed for each case by our experts to add more HPO terms prior to the exome analysis. Now, the quality of the data produced for clinical exome tests is really the key to an accurate result, and so in the next few slides, I'm going to review Invitae's exome assay performance with respect to detection of single nucleotide variants, indels, and copy number variation. First, I'm going to briefly discuss the basic steps to a next-generation sequencing assay. First, the genomic DNA is extracted from the patient sample. That DNA is then fragmented into smaller pieces, and we can use capture probes or baits to bind and catch the exonic DNA. That captured DNA is then sequenced, and the outcome is a bunch of small reads or copies of that DNA. These reads are then bioinformatically aligned to a reference sequence, and the variants are called. These detected variants can then be interpreted, classified, and finally, when appropriate, reported to the provider. Next slide. At Invitae, our exome assay offers comprehensive coverage of over 20,000 genes. We also have enhanced coverage of over 7,000 disease-associated genes, and 99.4% of the reportable range, which is generally considered to be exons plus or minus 10 base pairs, is covered at 20x. The average coverage is 150x. In terms of the types of variants that we can detect, we can detect, of course, single nucleotide variants, small deletions and insertions, less than 50 base pairs, and copy number variants may also be detected with high sensitivity, greater than or equal to four exons. Although I do want to say smaller CNVs, even single-exon CNVs, may also be detected and reported. Next slide. Importantly, the turnaround time for our test is six to eight weeks, and 95% of our cases are currently being reported on time. Invitae's exome test also allows for sensitive detection of intragenic copy number variation by next-generation sequencing. So the uniformity of coverage in next-generation sequencing tests is very important when it comes to designing high-quality assays. We know that the read depth can vary among exonic regions, but it can be reproducible amongst samples. We can see this in the example below, where each line represents a sequencing depth profile for an individual sample. We see that there's variation between coverage from exon to exon, but we can also clearly see a reproducible pattern of this coverage between samples. This uniformity that we see here is key to allowing for high-resolution intragenic CNV calling, and that's made possible by Invitae's state-of-the-art lab automation efforts. By testing many different samples, you can start to see patterns and create a baseline of depth that you normally would observe in any region, and then we can look for deviations with respect to the baseline samples. If in any particular example, for example, you see more reads or depth of coverage than you normally would observe, then that could mean that there is a duplication in that patient sample, and conversely, if you see less reads or coverage in that region, that may indicate that a patient has a deletion in that part of the genome. So by using this technology, exon sequencing can simultaneously identify both sequence variants and intragenic CNVs with high sensitivity from a single sample. In the next few slides, I'm going to describe how Invitae's moon utilizes data from Sherlock, our validated and quantitative variant classification platform. So as I discussed in previous slides, there are many challenges in germline genetic variant interpretation. Like many other labs, Invitae adheres to the framework put forth by the ACNG and the AMP. However, there is persistent concern of discordance between and within clinical laboratories, resulting from things like inappropriate application of variant interpretation rules, inconsistent use of evidence among scientists and laboratories, and finally, insufficient granularity for categorizing the evidence. Sherlock was created to try to address some of the issues that exist in interpreting germline genetic variants. Sherlock utilizes a linear process of variant interpretation that involves highly trained scientists, genetic counselors, and laboratory directors. It is a validated and quantitative variant interpretation method that ensures transparency and consistency, and it also occurs in an auditable, custom-built software platform. Sherlock was created on the bedrock of the 28 rules that were established by the ACNG and AMP guidelines. In the course of over 7 years and 15 revisions, after interpreting variants in over 11,000 genes and interpreting over 790,000 interpreted variants, we have created over 200 detailed refinements to the ACNG and AMP rules. Sherlock considers five types of evidence in a hierarchical approach. The first is population data. The second is variant type. The third is clinical observations, and these are both clinical observations that we see internally during testing at Invitae and also externally in the literature. The fourth are experimental studies, and the fifth are indirect and computational evidence. We utilize a rule-based scoring system for each individual piece of evidence, using either benign or pathogenic points, and these points are applied to a point-based score threshold to determine the final classification based on the ACNG-suggested five-tier classification system. So next, I want to provide you with a little bit more detail about the types of evidence that we use in variant interpretation. So for population data, you can see that benign or pathogenic points are awarded depending on the frequency or the rarity of the variant in population databases. We also take a look at sequence observation. So the different types of variants that we see, whether the variant is missense or perhaps a predicted loss of function variant. We also take a look at various types of clinical findings. So was the variant determined to be de novo? Do we see the variant in individuals that have very specific phenotypes for a gene and disorder in question? If the variant occurs in a recessive gene, do we see maybe two variants that are in trans? And then we also take a look at segregation data. In terms of molecular studies, we do a thorough and rigorous analysis of the studies. And depending on the type of study that is performed and the evidence, we can apply either strong or weak evidence towards the variant either being neutral or causing a disrupted protein product. Finally, we take a look at computationally predictive evidence, trying to determine if the evidence points towards the variant being neutral or deleterious. In terms of evaluating molecular studies and computational predictions, one difference at Envitae is that we use a validated functional modeling tool to better inform these criteria for missense variants specifically. Envitae's functional modeling platform or FMP tools use training data to create gene-specific models that can accurately inform whether specific missense variants are pathogenic. So for in silico tools, we have sequence and biophysical analyses. For indirect evidence, we can use spatial analyses. And for experimental evidence, we have molecular and cellular assays. All of these together allow us to test hypotheses about the impact of missense variants on both protein structure and function. I do wanna note here that unlike existing computational approaches that have a single model for all genes, Envitae's functional modeling platform does train to evaluate variants in each gene separately. And that allows us to take into account disease-specific domains and molecular mechanisms, allowing for a much more accurate classification. One of the things that makes Envitae's functional modeling platform so powerful is that we have instituted prospective performance of our FMP tools. Using unknown variants. So I'm gonna show two examples here. In example A, the variant was entered into ClinVar as a variant of uncertain significance in 2018. Later that same year, our FMP tool predicted this particular variant to be damaging. And in 2019, another independent laboratory classified this variant as pathogenic in ClinVar. So this essentially supports that our predicted damaging evidence generated by our FMP tool was accurately predicting the variant to be deleterious. The example here in variant B, in 2018, the variant was not present in ClinVar, but our FMP tool predicted that the variant would be benign. And then in 2019, an independent laboratory classified the variant as benign in ClinVar. Again, giving us confidence that this tool is accurately predicting the variant to be benign. Importantly, in a set of over 700 variants for which our FMP tools predicted either damaging or benign evidence, and an independent classification of the variant was later uploaded into ClinVar, we found that our tools accurately predicted damaging or benign evidence over 95% of the time. And because of this, we use prospective performance to test all of our FMP tools. And this is an important part of the assessment for when a tool meets our accuracy standard and can be used in clinical variant interpretation. Envitae's variant interpretation process considers information from both public databases like ClinVar, for example, and also Envitae's internal database. Importantly, at Envitae, over 1 million tests have been performed in many different disease areas. And we use this wealth of knowledge to provide more accurate variant interpretations for all of our patients. We also believe that at Envitae that data sharing is critical among laboratories. For those that aren't familiar, ClinVar is a freely accessible public archive of reports of the relationships between human variation of phenotypes with supporting evidence. And as of August of 2020, Envitae was the top submitter to ClinVar with over 392,000 submissions. So now I'd like to switch gears a little bit and talk more about Envitae's intelligent engine, MOON, and how it works in combination with phenotypic information to return a list of suspicious variants for expert review. Recently, Envitae acquired Diploid, the maker of MOON and Apollo software. In the next few slides, I'm going to discuss some exciting enhancements to our exome and assay as we integrate these softwares into our exome analysis platform. The MOON platform has a proven history of providing rapid and accurate results. MOON has been reported in multiple manuscripts, including studies from Rady Children's Hospital, where they utilize MOON to interpret whole genome sequencing data from patients in the NICU. MOON was shown to find 99% of the causative pathogenic variants identified via expert manual review in 95 children. This software was also instrumental in Rady's Guinness World Record for genome analysis. MOON is an intelligent engine that integrates and weighs genetic and phenotypic data to identify suspicious variants. Apollo offers continuous weighted gene phenotype curation from scientific literature. And in a pilot study of 150 previously solved Envitae exome cases, MOON alone correctly identified 97% of the causative variants in less than two minutes per case. So much like Rady's experience, Envitae also found high precision compared to expert manual analysis. Next, I would like to describe how the MOON platform works. Input into MOON is raw data from the patient in the form of variant call files, any HPO terms describing the patient's phenotype, the patient's biological sex and age of onset of features, and also information about the family, including whether there are other affected individuals in the family. What MOON does is weigh variants based on public data sources like NOMAD, ClinVar, and dbSNP, internal data sources, for example, variants that are previously interpreted at Envitae or Sherlock variant interpretation framework criteria. And it also considers information from Apollo, which again is a gene disease relationship network curated daily from a combination of automated literature searches and manual expert quality control of gene disease relationships. So the result here is an intelligent engine that integrates in ways molecular effect, scientific evidence, and clinical phenotype to identify suspicious variants for a case. Apollo, again, is a continuous weighted gene phenotype curation that occurs from the scientific literature. So the literature is continuously searched and through natural language processing, HPO terms are extracted from the literature. There is then a manual curation and quality control of the findings. So for every gene disease association, phenotypes are listed, but also ranked based on the frequency in which that feature is associated with the gene disease and the literature. We are also working to include our own internal cases to further refine the databases. This happens real time with the emergence of new literature and therefore provides a real time meaningful gene disease curated database. Currently, the Apollo database has quite high quality phenotypic information for more than 6,000 gene disease relationships. This ever growing knowledge base integrates new scientific insights in real time, which guarantees a comprehensive and up-to-date analysis for every patient. One of the most important parts of Invitae's exome sequencing interpretation is our clinical team and the expert review that they provide for each and every case. In the next few slides, I'm going to talk about our expert team and also the results that you can expect from obtaining exome sequencing at Invitae. Invitae's clinical team of exome experts is comprised of an experienced team of PhD scientists, certified genetic counselors, and American Board of Medical Genetics and Genomics Certified Laboratory Directors. We have broad experience in clinical and academic laboratories. And importantly, we have sequenced several thousand probands and over 10,000 individuals who have been sequenced via our exome test. Our team of exome experts aims to provide a personalized report with clinically relevant variants. To help us, Moon provides an objective list of suspicious variants for manual review. And Sherlock enables accurate classification of the reported variants. Our reports include pathogenic variants, likely pathogenic variants, and variants of uncertain significance in genes when phenotypes and inheritance pattern match the known gene disease associations. We also report single pathogenic and likely pathogenic variants in autosomal recessive genes when there's a strong clinical overlap with the patient's phenotype, treatment is available, or when further testing might rule the condition in or out. Overall, our goal is to identify a variant in a gene or genes with strong clinical overlap to the patient's indication for testing. So now we've reached perhaps the most exciting part of the talk today, where we're going to provide an overview of NVTA's re-evaluation and re-analysis programs. There have been several recent studies that have shown the importance of case level exome re-analysis. And the ACMG recently issued a statement outlining points to consider in the re-evaluation and re-analysis of genomic test results. Following the adoption of clinical next generation in exome sequencing, the rate of discovery of novel disease gene associations skyrocketed in the early 2000s. And while the rate of novel disease gene discovery has slowed in recent years, it still remains over 500 genes per year. And what I would like everyone to consider here is that anytime a patient is sequenced and then years go by where many different new novel disease genes are discovered, that sort of represents a missed opportunity for that patient to achieve a molecular diagnosis. So it's very important that exome data be reconsidered over time. Before I move on to talk about NVTA's re-evaluation and re-analysis programs, I would like to provide clarification of these two terms. So the first term, variant level re-evaluation, occurs when a report is, updated report is issued because of known variant upgrades or downgrades, or when a gene is now considered a disease gene, whereas before it was considered a candidate gene. These types of re-evaluation cases are identified through internal databases based on changes to a specific variant or gene, and then the corresponding cases are reviewed and reports are updated. So this is a pretty narrow focus, the variant level re-evaluation. On the other hand, case level re-analysis is a re-review of all variants that were identified in a case, both reported and unreported in the context of the patient's phenotype. And this can also include maybe new or updated phenotypic information that wasn't available at the time of the first analysis. Many labs do state that they do a continuous re-analysis in the background, but really what they're talking about here is variant level re-evaluation, not really case level re-analysis like we're talking about here. Case level re-analysis has demonstrated to lead to an increased yield of up to 20% with the highest yields achieved over one year after the first analysis. So at Invitae, we are committed to achieving high diagnostic yields for all of our patients. And so we have instituted both re-evaluation and re-analysis policies that I'd like to share with you today. The first variant level re-evaluation is something that Invitae has always done and will continue to do where variants are re-evaluated when new data become available. The second case level re-analysis is something, a new policy that I'm introducing to everybody today. And this is an expert case level re-analysis that can include any new clinical information that you provide every six months for a minimum of three years. When we do a case level re-analysis, any new positive findings are reported and clinician requested re-analysis may also be available. This expert case level re-analysis that occurs on a six month cadence puts less burden on the patient and provider. So there is no need to remember to request a re-analysis or decide on the best time to ask for one. This ensures that patients receive the most up-to-date information about their genetic data. Case level re-analysis is offered at no additional charge as it is included as part of the original order of the test. So next I wanted to address how Invitae is addressing some of the challenges that we talked about in the beginning of the talk. So first you might remember we talked about this constant evolution of gene phenotype information. We are tackling this issue by using Apollo, which again is a continuous weighted gene phenotype curation that is occurring from scientific literature. Natural language processing is used to scan the literature and importantly it's followed by manual expert quality control of gene disease relationships. We also talked about difficulty in performing a reproducible and unbiased evaluation of genetic variation in the context of our probands phenotype. And today we presented Invitae's mood where we integrate and weigh information like molecular effects, scientific evidence and the patient's clinical phenotype to identify suspicious variants for each case. We also talked about the need for a consistent evidence-based gene disease curation and variant interpretation framework. And I explained today about Invitae's use of Sherlock which is our automated consistent and comprehensive refinement of the ACMG AMP variant classification criteria. And finally we talked about performing reanalysis at scale and I just presented to you Invitae's policy of offering expert case level reanalysis for all patients every six months.

Invitae

exome platform

affordable

genetic diagnosis

challenges of exome sequencing

gene-disease curation

variant interpretation frameworks