The below text is a transcript from the webinar. Because it is a transcript, there may be oddities that arise from the process of translating speech into text. We recommend accessing the recording, above, to gain full context.
Broader adoption of next-generation sequencing-based cancer testing reveals the challenges associated with variant interpretation and underscores the need to standardize results across molecular pathology labs. Guidelines developed by the Association for Molecular Pathology (AMP) with representation from key institutions, including ASCO, CAP, and ACMG, outline a system to categorize sequence variations in cancer.
In this webinar recording, Medical Director, Shalini Verma, and VP Product, Andy Bredemeyer, review the guidelines from AMP and demonstrate how to put them into practice using real clinical examples.
As you are aware that recently, CAP and ASCO published joint consensus recommendations on the four clinical interpretations of somatic variants. This article provides guidance on various aspects of molecular oncology case reporting, to results reporting. For today's discussion, I will focus on the various aspects of somatic variant interpretation, how we apply these recommendations in our clinical practice, and some additional key points from the guidelines that we believe we should keep in mind when evaluating the results of NGS-based cancer tests. My colleague Dr. Andy Bredemeyer, will go over the clinical genomics workspace, or CGW, software application. That is utilized in many laboratories today across the country and internationally, for streamlining the clinical ideas testing workflow.
Before I dive deep into the various aspects of somatic variant clinical interpretation, I would like to take a moment to orient us all to the NGS testing workflow for cancer samples. So it starts with receipt of a tissue sample from a patient with malignant cells in it in a clinical NGS laboratory, where a sample is received, nucleic acid is extracted, library prep is done, and sequencing is performed on the nucleic acid. Once that's performed, then what starts is the dry bench, or the bioinformatic process, and then finally the interpretation section report sign out.
Then comes the primary, secondary, and tertiary analysis, the process of evaluating and analyzing data parameters, determining what are the real and reportable variants, which we call analytic interpretation, and then determining what do these real or reportable variants results mean for a patient's specific situation, or clinical situation, which as we also call clinical interpretation. Then, in the end, a molecular diagnostic professional finalizes the report and signs off on it, and it becomes the part of the electronic medical record system for our patients, and is utilized towards making decisions for good treatment and management.
I would again like to point out two key points, here. That, in the process of performing interpretation there are really actually two types of interpretation that were performed. One is called analytic interpretation where one examines the raw data, determines the quality/quantity of the analyte, or the variants, in this case, and determines what is a reportable result. The second part of it is the clinical interpretation, where one determines what do these reportable variants mean for a patient with respect to their disease and clinical situation. And this is what I will be focusing on today, over the next couple of slides, with reference to all the guidelines and the recommendations on how we should be performing this process of clinical interpretation for somatic variants.
The AMP/CAP/ASCO guidelines underscore the need to focus the interpretation of somatic variants on the impact on clinical care. And towards this end, the guidelines in the published literature recommends that when one is performing clinical interpretation and classification of the sequence variant in cancer, one should collect evidence for variants' clinical significance, determine its clinical impact for the patient, with respect to diagnosis, prognosis, therapy, and preventative action, and evaluate the strength of the evidence for clinical impact.
So what does it mean to collect evidence? So, the clinical evidence for a genetic alteration is of various types, and should be collected from different, multiple sources for these recommendations. So briefly, what that looks like is that one should be evaluating the mutation type of a genetic alteration at hand, whether it's an activating mutation, whether it's a loss of function mutation, is it a carbon number variation, or a fusion? Is it present or absent in various somatic variant population databases? What is the variant allele frequency of the genetic alteration, could it be a germline variant? If it has a 50% or 100% MAF, what is the minor allele frequency or the population frequency of this variant, whether it's ... what is the functional characterization of this variant? Has anybody described in functional studies and population studies, or there are any other prediction algorithms that help us understand the functionality of this mutation. And if this variant is important in any particular biological pathway.
One is also recommended to review certain information sources to assess the clinical impact of the genetic alteration of the variant at hand. These are FDA approved therapies, professional guidelines, well-powered studies with consensus, investigational therapies including clinical trials, small studies with and without consensus, case reports, and pre-clinical studies. So once we have collected, for every variant in our patient, all different types of evidence from all different sources, then what we need to do next is, for these recommendations, we have to assess the clinical impact of that variant. Is it of diagnostic, prognostic, therapeutic, or preventative significance?
Once that's established, then one is also required, or recommended, to evaluate the strength of that clinical evidence, or experimental evidence. And the join consensus recommends classifying this evidence as level A, B, C, and D, depending on from which source we have acquired it. So briefly, if the source is FDA approved therapies and professional guidelines for this clinical evidence, then it's a level A strength. If the sources are well-powered sources with consensus from experts with the field, it's level B. Multiple small studies with some consensus and clinical trials, the evidence from these sources is categorized as level C, and if the evidence is coming from preclinical studies or some small studies or few case reports without consensus, then it is categorized as level D.
So with that, moving on to the next slide. So one we have collected all different types of evidence from all different appropriate sources and we have evaluated the clinical impact of this evidence in context of patient's disease and other diseases. And we have evaluated the strength of the clinical evidence based on these sources. Then one is asked or recommended to perform the evidence-based categorization or classification of the somatic variants. So what does this look like based on the joint consensus recommendations? So tier one are variants of strong clinical significance. The clinical significance, I would like to remind, could be of therapeutic, prognostic, or diagnostic significance, or even preventative. And the evidence letter that that fits in this tier is evidence level A and B. Tier two is all variants of potential clinical significance, and the evidence level that supports this tier is level C and D. Tier three is variants of unknown clinical significance, and tier four are benign, or likely benign variants.
So how do we apply this when we are actually performing clinical interpretation and what it looks like. So, when we are drafting a clinical interpretation for somatic variant, again I would like us to remember, and would like to remind us that interpretation of somatic variants should be focused on the impact on clinical care. So with that said, the first thing one needs to do is look up the pathological diagnosis of the patient. With that I mean, if there's a pathology report, understand what's the site, diagnosis, subtypes, stage, grade of the disease for the patient.
In case that is not available for some reason, example, there's an occult primary, then one must consider, or try to understand, what is the intended use for these test results, so that you can draft those interpretations in a really meaningful manner for the patient. Once that's accomplished, then we have to go through that process that's recommended in the guidelines, which is to understand from various different sources and collect the evidence about the gene, the variant, like I mentioned earlier, the mutation type, variant allele frequency, population frequency, presence or absence of that genetic alteration in various databases, its involvement in any biological pathway, and its functional characterization based on the evidence in published functional and population studies and from prediction algorithms, for reference only.
Then, we again assess the clinical impact of this genetic alteration, based on the evidence available in different sources, as mentioned earlier, as the approved drug labels, professional guidelines, well-powered studies with consensus, clinical trials, small studies with and without consensus, case reports, pre-clinical studies. And one should also consider the role of co-occurring variants in context of patient's diagnosis. If two variants or three variants together have a diagnostic, prognostic, or therapeutic relevance for patients' disease, then we perform the evidence-based categorization and classification of the variants, as recommended by the joint working group. That is the variants of strong or potential clinical significance, VUSs, and benign or likely benign polymorphisms. If it is necessary, or if it is determined to be important, then we also are encouraged to provide additional recommendations, which if are provided, then they should be short, worded carefully, and they should be based on evidence, with appropriate literature citations. There are some additional points that I would like to highlight here and share with you regarding the reporting of clinical significance. So the guidelines say that for tier one and two variants, that is, the variants of strong and potential clinical significance, it is really important to remember that these alterations should be interpreted in the clinical pathological context. That is, in context of patients' disease. They also say and state that, do not make specific clinical trial recommendations, but it is okay to make general statements about availability of relevant trials, or citing results of published trials is also acceptable. When we are mentioning VUSs or variants of unknown significance, or tier three variants, the recommendation is that we provide the most critical information in the reports, which should be presented in a very concise, clear manner. And the comments could include descriptions of function, prognostic, or predictive significance of the variants for the particular tumor type, its impact on biochemical pathways, and its prevalence in relevant cancers.
The other point that I'm going to highlight over here is really, really important to remember when drafting these clinical interpretations. And that is that one should remember that the treatment and management decisions, for any patient, are based on many pieces of medical information. The suitability for treatment is based on many factors, other than the diagnosis as we find it written on the requisition form, and the genotype that was discovered through testing. And most often, and very often, these factors are unknown to the molecular professionals who are reporting these results. And a failure to take these factors into consideration when recommending a specific therapy, can lead to confusion, conflict between patient and oncology team, and a lot of anxiety. So these treatment suggestions, the guidelines say, should be evidence-based, they should be relevant to patient's diagnosis, and in the report it must be clearly stated that this report contains only generalized treatment suggestions, incorporating the data points that are relevant to the laboratory, the diagnosis as written on the requisition form, and genotype. But additional factors need to be incorporated into crafting a treatment plan for each individual.
So I will also review what I just discussed and shared with you in form of two examples. So first case example here is a 57-year-old man who was diagnosed with non-small cell lung cancer and NGS testing revealed a MET variant. And again, keeping in mind that interpretation of the somatic variants should be focused on their impact on clinical care. So, we went through the process as recommended in the guidelines of collecting the evidence. What we found out when we were reviewing those different literature sources that MET was a potential therapeutic target in a number of cancers, including non-small cell lung cancer, and NCCN guidelines state that Met exon 14 skipping mutations are of clinical relevance, actually therapeutic significance, in non-small cell lung cancer.
So further analysis of this variant, we found that this is a 12 base pair deletion, in MET intron 13, and it is located 15 base pair upstream of five prime end of MET exon 14. This variant will be found that was never reported in literature, this was not characterized as a splice site variant in the literature, and it was not characterized for its function consequence on MET exon 14 skipping. The variant allele frequency for this variant was 20% in patient sample, the population frequency in all databases was less than one percent. While reviewing the literature what we also found was that other EMT intron 13 deletions that were overlapping the region of this variant that was found in our base and had been reported in lung adenocarcinomas and some other tumors in two case studies. And in some of these cases, MET inhibitors like crizotinib had been tried for treatment. We also evaluated different literature sources for clinical impact, and we could not find any clinical significance information available for this variant.
So based on all the evidence, based on assessment of clinical impact, and the strength of the evidence, we classified this variant as tier three, a variant of uncertain significance. And in our recommendations, we also mentioned, based on the evidence that we found, that since MET intron 13 deletions overlapping the region of patient's variant have been reported in lung adenocarcinomas and some other tumors in two case studies, and in some of these cases since MET inhibitors were tried for treatment, clinical correlation is recommended.
The next example here that I would like to share is from a female patient, 53-year-old woman that was diagnosed with hairy cell leukemia, and the NGS testing in the basis tumor sample, which was a bone marrow, revealed a BRAF mutation. So, BRAF, V600E was the variant that was discovered. And again, keeping in mind and emphasizing that the interpretation of the somatic variants should be focused on their impact on clinical care. The evidence that we collected is revealed that BRAF is an oncogene, involved in RAS/MAPK pathway. This is a missense mutation with a variant allele frequency of 60% in patient's bone marrow sample, and it's present at less than one percent population frequency in all databases. This is also one of the most frequent BRAF mutations that has been identified in human cancers, and occurs in virtually all hairy cell leukemia patients.
With respect to functionality of the mutation, we found that V600E occurs in the kinase domain of BRAF G, and causes activation of the RAF/MEK/ERK signaling pathway. With respect to clinical impact, the guidelines in the published literature stated that BRAF V600E has been reported in majority of the patients with classic hairy cell leukemia. It is useful for distinction of this disease from hairy cell leukemia-variant and other splenic B-cell lymphomas. Vemurafenib, a BRAF V600E kinase inhibitor, has also demonstrated activity in relapsed or refractory hairy cell leukemia with V500E mutation.
So based on the guidelines and the published literature, and NCCN data, we established that this variant is of diagnostic and therapeutic significance in hairy cell leukemia. So we, based on all the evidence and the strength of the evidence, we classified it as tier one, variant of strong clinical significance, with no additional recommendations.
So with that, I will move into the next section of my talk, where I would like to share, as I mentioned, some additional key points that we believe one should keep in mind when performing the interpretation for these somatic variants. So the first topic here is polymorphisms. So I would like to start with, do not miss variants which have more than one percent or one percent population frequency, and may be of clinical significance for your patient. An example of such variants is pathogenic variants in genes associated cancer predisposition syndrome, activating mutation with predictive or prognostic impact, while they all have more than one percent or one percent population frequency.
Then, a few points about polymorphisms. For the guidelines, the variants that we call as polymorphisms, for them, no existing published evidence of cancer association. There is no existing published evidence of cancer association. These are benign or likely benign variants, and they're observed at significant allele frequency in the general or specific population databases. These guidelines also recommend to not report these tier four alterations of benign or likely benign variants in the report. It is again, really, really important to remember that one really must pay a lot of attention for these variants that have more than one percent MAF, by reviewing disease mutation databases like ClinVar and literature for assertion of pathogenicity and clinical significance.
And I would like to conclude this discussion on polymorphisms by sharing this example, of KDR Q472H that we see all the time in our patient population. This is a known polymorphism with very high population frequencies, but it is also an activating mutation. There's recent literature that establishes that this is a pathogenic germline variant in melanoma, and melanoma patient with this alteration may benefit from anti-angiogenesis treatment.
And towards this, there are some other key points that I'd like to bring to your attention. So first is minor allele frequency. SO it is really important to know and remember that there is no standardized cutoff for MAF that could be used for eliminating polymorphic or benign variants. In absence of the paired normal tissue, the AMP/ASCO/CAP work group recommends using one percent as a primary cutoff. Again, aggregate global mass is commonly used in laboratories, and one may consider using ethnicity-specific MAFs based on the ethnic background of the patient, if that is known to you.
This work group also cautions about the use of population databases to exclude polymorphisms, and the reason that they state is, one, that several well-known classic cancer-associated and target table somatic alterations are included as germline variants in population databases. And the reason for that could be because the individuals who are participating in these sequencing studies that have made it into these databases were assumed to be healthy or free of supplemental disease at the time of participation in the study.
The second topic along this line that I would like to discuss is germline pathogenic variants. So tumor sequencing with or without matched normal tissues may reveal variants that are of germline origin. In this regard, again, AMP/ASCO/CAP joint consensus recommends that one should report germline variants with known evidence of clinical impact. And one should also report germline pathogenic variants in genes that are associated with hereditary cancer syndrome, that has an established guideline for clinical surveillance, and one should do this with recommendation for cancer genetic counseling. Some additional guidance on this topic is that it is really, really important to distinguish somatic variants from inherited germline variants. Clinical laboratories must have polices that address detection, disclosure, non-disclosure, interpretation, and reporting of germline variants.
If one is interpreting and reporting germline variants, one must follow the ACMG standards and guidelines for such interpretations. Whenever a pathogenic germline variant is suspected during tumor-only testing, confirmation of the variant with the normal tissue sample, along with appropriate genetic counseling, should be recommended. Laboratories should also have policy about testing germline sample for a variant found in a malignancy, to confirm germline or somatic origin. And these policies should state the use of clinically validated germline test, after appropriate patient consent is received, or per request of a clinician. If secondary findings are revealed in germline testing, the ACMG recommends disclosure of the positive germline results for the 53 genes. This disclosure is also recommended even when the germline variant is only being evaluated as part of a tumor/normal study. So the guidelines say that by inference, it seems prudent that one should consider the likelihood of germline pathogenic variants in a tumor-only somatic mutation study. And I would like to end the discussion on this topic by stating that germline variants may also serve as an inclusion criteria on clinical trials.
Now briefly, about assessment of mutation functions when we are collecting that evidence for our genetic variant on hand. So the guidelines state that result of prediction algorithms should never be used as the sole evidence for variant classification and clinical decision-making. One is encouraged to utilize data with consensus from published functional studies, population studies, and potential drug response studies, to determine functionality of a mutation. In this regard it is also very important to remember and keep in mind that in general, missense and splice site prediction tools have a moderate specificity of 60% to 80%. And they have a tendency of over-predicting the deleterious impact of the alterations. Thus, one should exercise caution when interpreting in silico scores and this information is to be used for reference only, and that has been stated multiple times during, if you read that article on the somatic variants interpretation.
So, some additional points to remember when reporting the somatic variants for patients. These detected variants should be carefully reviewed by appropriately trained and certified molecular diagnostic professionals in the context of each complete case. Histologic findings and evidence-based variant categorization, as we discussed earlier, must be performed before reporting. The work group also recommends that these reports should be short, simple, and to the point. Tier one to three must be reported in descending order of clinical importance. And again, as I mentioned earlier, it is not recommended to include tier four, or benign or likely benign alterations in the report. All detected genetic alterations should be reported in standard nomenclature, in addition to the colloquial nomenclature so that it's clearly communicated to the oncologist and the treating physician.
The recommendations also state that pertinent negatives should be reported in a disease-specific manner, and they should be included for tier one drug/cancer combinations. And what that really means is, in patients with lung cancer, a definitive lack of an EGFR mutation must be stated. In a patient with melanoma, a definitive lack of BRAF mutation must be stated. If the germline variants are not reported in some of the genes in an NGS panel in a laboratory, than the initial report should specifically state that fact. Additionally, if the NGS test does not allow definitive differentiation between the germline and somatic variants, the clinical report should also clearly state that. If there's any uncertainty about the patients results due to issues like sequence quality, sample adequacy, tumor content, lack of biomedical knowledge, then that must be clearly reported or communicated in these molecular reports.
So with that, I would say that so far, what I discussed was the process of performing clinical interpretations for somatic variants based on these recommendations. And what does it look like if we were to do this manually/ one would be opening all these windows on their computer or visiting all these websites and all these data sources and many more to collect information on every single variant for their patient, and collate that information and then try to assign clinical meaning to it before they sign out the report. And that is a daunting task, especially when you have a lot of variants on hand that need to be interpreted and reported.
So towards that, it has already been recognized in literature and has been reported in literature for last many years at least since 2013, that with hundreds to thousands of tumor variants observed in the coding region of an individual's genome, it will not be possible for molecular pathologists or oncologists to identify and appropriately annotate the clinical significance of each variant manually. So clearly there is a need for prioritization strategies for identification and reporting of clinically significant genetic variants. It has been also recognized that once one has narrowed the list of candidate variants, the biggest challenge that still remains on hand is to interpret the remaining genetic alterations within a biological context. Even today, the clinical interpretation of most of the variants identified in NGS-based cancer diagnostic tests usually involves a manual search of published literature like I shared in the previous slides. And it is a burdensome process and a daunting challenge.
So, the problem has been identified and there have been many discussions about how one could overcome this challenge. So these discussions I believe, resulted in this [inaudible 00:35:46] paper in 2014 from Harvard that talks about creating a clinical genomics work station. And I'll briefly review the features of this genomics work station that would help in streamlining an NGS workflow in a clinical laboratory. So they state that this work station would have two big components. One is a genomic information management software which is integrated with the medical knowledge base.
So briefly, this genomic information management software would have computation tools that support the rigorous analysis and clinical interpretation of comprehensive genomic data. It would assist in determining real and reportable variants that does help us in performing analytic interpretation, help us prioritizing the non-synonymous mutations over the synonymous and intronic mutations, and recognize cancer mutations that have been previously annotated and reported, and would be dynamic in nature. The knowledge base with which this information management software would be integrated would contain curated information stored for future use and would comprise of previously classified and interpreted variants, curated annotations of most frequent and relevant variants and multiple tumor types based on publicly available resources. It would be designed to accommodate new clinical pre-clinical data, and would be routinely updated with new scientific and clinical knowledge as it becomes available.
This medical knowledge base would be integrated with the genomics information management software for automated report generations. And if one such genomics work station were to exist, it would, I believe, alleviate at least 60 to 70% of the manual work that we envision in that process of performing clinical interpretation and reporting of the variants for a clinical NGS case.
So here at PierianDx we have developed one such clinical genomics work station and we call it clinical genomics workspace or CGW. And on a lighter note, I would like to point out and say that this CGW was developed at Washington University with some of our founders in 2011, 2012. My colleague Dr. Andy Bredemeyer in the next few slides, in the next section of the talk, will review the features and functionalities of this CGW, how it has been implemented across many large cancer centers across United States and internationally, and how it is helping out several clinical and NGS labs in streamlining their NGS workflow.
I'd like to go into greater depth in a single example to demonstrate how a clinical genetics software platform can aid in the evaluation of variants, their clinical impact, determining that strength of clinical evidence, and classifying and interpreting variants.
The goal of Clinical Genomics Workspace (CGW) and our knowledgebase is to streamline the workflow of clinical NGS testing, and to accelerate it so that you can issue high-quality molecular genomic reports to an ordering physician and their patient with quick turn around time. And core to this capability is the knowledge base that we offer. This knowledge base is by no means static, we are updating it weekly, we're adding to it and expanding it all the time, based on new knowledge. Today it covers approximately 1100 genes that have been demonstrated to be relevant to cancer, as somatic cancer. We cover about 6 megabases of sequence. And this is to point out that we not only match content from our knowledge base to specific variants or specific syntax to variants, but also across ranges of genomic space, different variant types, and variant consequences. And so that allows us to match content to trickier variant types like copy number changes and fusions.
A key feature of our knowledge base is that it includes the largest clinical genomic sharing network available. I think of it in a couple of components. One is the shared interpretation knowledge base, and this is composed of interpretations written by users on our platform, as well as or own interpretation services team. Clinical interpretations that have been put into reports and signed out and issued to patients. And so you can leverage these interpretations for a kickstart to your own reporting and interpretation drafting, you can glean information from them if they're synthesized sets of information from experts. And it also helps to automatically classify variants when there may not be information known about these variants from practice guidelines or drug labels. And the other component of this shared knowledge base is shared, de-identified, aggregate variant data. And from this you can learn about the frequency with which certain variants are detected across different cancer types, across this whole network of users on our platform. Likewise, you can learn about how users on our platform have classified such variants.
Remaining content in our knowledgebase includes public sources we have aggregated and loaded to our system as well as some key clinical content that we have curated. Our own annotation team has taken FDA drug labels, practice guidelines from ASCO and NCCN, as well as clinical trials, and this content has been read through, and carefully turned into discrete logic so that specific inferences can be made on variants and patient cases, that is, drug responsiveness, prognostic or diagnostic information. And so, this is automatically shown to you in the system, so you can decide how that impacts your evaluation of the strength of clinical evidence.
So the example case I'll describe is a non-small cell lung cancer patient, a 72 year old male, who had previously had an ALK rearrangement identified. And they were placed on ALK inhibitors, but the disease progressed after a time. A new biopsy was taken and sequenced again, and her are some key variants that fell out of the analysis. The ALK fusion was again identified, as well as an ALK, single nucleotide variant, a G1202R. And then two TP53 variants, G154V, and P72R. So at a glance this ALK single nucleotide variant has been described as secondary mutation that confers resistance to ALK inhibitors, first and second generation of drugs. And this patient has progressed after treatment with those inhibitors, so that appears to be a reasonable hypothesis. Also, TP53 mutations in lung cancer have been reported to be negative prognostic factors and can also serve as eligibility criteria for trials. A point of interest is that a similar such case was published in CAP today, earlier this year, an ALK rearranged lung cancer and a subsequent escaped mutant. So that's something you might look into for more information on such an example.
Now, the initial step before one can perform that clinical impact assessment on variants is to decide which variants are really worth looking at, into in more detail. And some are quite large, and you don't start with a list of 200 variants and assess them all for clinical impact, you need a strategy for narrowing that down. So after determining that you have the sequence quality, of your case is high quality, and we've performed that sort of BAM file based assessment, or BSQ file based assessment for quality, and you determine that the sequence quality is high, you'd proceed to looking at the list of variants, the return from the secondary analysis, and deciding which of these will warrant further investigation.
And so in our variant table, or variant tab view shown here, you can see that you can assess both basic annotations on the called variants, that is, what is the gene in which the variant falls, what is the syntax of the variant? The consequence of that variant, is it a synonymous change or a non-synonymous change? What is the variant type? As well as then assertions about the variant from VCF itself, so what is the variant allele fraction, the depth strand bias values, et cetera. You'll also see that the variants are classified automatically across the classification scheme in our system based on content in our knowledge base.
Also, notably, as you perform a validation and gather data, or across many clinical cases, you may determine, if you develop that variant database, that there's some recurring variants that have not been described as relevant to cancer, and you might upon investigation find that they are low-quality variants that are indeed artifact. And our system allows you to flag such artifacts, such that they are automatically left off of future case's reports. You can always view these anyhow and retrieve them if you deem necessary on a given case, but this is a good way to identify a false positive and prevent them from being recorded in the future.
A key piece of determining which variants need to be assessed for clinical impact is to develop a standardized, routinized method for filtering variants. In this example, showing up in the green box, I've applied a relatively simple filter, it's a subset of the total variants called in the VCS. That filter's based on automatic classification level assignment, from CGW, the consequence of the variant, looking at non-synonymous changes only, and then applying some basic quality criteria, like the variant allele fraction 5%, and making sure that the variants aren't strongly strand-biased. And these quality criteria that are applied in such a filter are configurable and you would determine, based on your validation data. Okay, so now let's look at a few of these specific variants that we focused in on. You'll have seen from that short list that we just looked at that these are the same that I called out initially in the framing up of the case. So from that view that I showed you of a variant table, one can link out to separate window for each specific variant in the case. We refer to this as our variant details page, and this shows all of the knowledge-based content that we have curated and aggregated from other sources that's relevant to this variant. So we've specifically shown content here that is relevant based on the patient's disease and the variant identified.
So for such well-established variants, like this ALK fusion, there are already drug labels available that we've curated, practice guidelines as well, and so our system allows you then, on our guidelines tab, of this variant details page, to quickly access the discrete information from those guidelines or drug label documents, to find the drugs to which a tumor of this variants might be expected to be responsive or non-responsive, or other diagnostic, prognostic information from those practice guidelines. Likewise, our interpretations tab in the variant details page gives you a view of all of the available interpretations that have been written by users of our platform, including our own interpretations services team, and then find out. So you can leverage these existing interpretations, use them in your report, edit them as you see fit, and then importantly, when you sign out your case, the edited interpretation that you've created for this variant will be saved and associated with that variant. And in future cases when you encounter the same variant and disease, you will find that your interpretation automatically populated your report. Moving to the ALK G1202R variant, this is less well characterized. There are certainly studies ongoing, clinical trials available, but there is no approved drug and there are no practice guidelines that have been issues for this variant. And so for these somewhat less well characterized variants, you can lean on our interpretation knowledge base. This has been encountered by other users of our platform, and the existing interpretations allow several key benefits. The first is that it will classify such variants that are not well characterized and not in practice guidelines across the tiering scheme to properly surface those variants that are of clinical meaning when you first interact with the automatically generated report in our system. Second, these interpretations have citations of publications and cite also trials that are maybe relevant, and so you can harvest that information and follow up on those publications to read more. And then, of course, you can use this interpretation as a starting point for your own interpretation. Change the style to your liking, add any newly available publications you might find, and sign out the report.
Clinical trials are available and matched within our system. We're curating these and updating them weekly from clinicaltrials.gov. You can use matching trials to place your variant into tier two with evidence level C across the joint consensus guidelines' recommended classification scheme. While they say not to recommend a specific trial, we do feel it's appropriate to list possible trials of interest, and certainly be able to make a statement about availability of trials.
The joint consensus guidelines also state that it's important to look at cancer variant databases, and so we've loaded up the cancer genome atlas, the CGA, as well as COSMIC, and keep that up to date, and you can link out from our human research evidence tab to these COSMIC pages. We've also run computational algorithms on these variants, so while the guidelines do state that you cannot use this as stand-alone evidence, it's some supporting information that is available to you on variants that have been called in cases on our platform.
And then when a variant may not necessarily have practice guidelines or interpretations available, the clinical evidence from databases is lacking, then we also make it possible to quickly search for literature that might ultimately help you make a determination on whether this is a VUS, or perhaps it would fall on level 2C or 2D, based on some more recently published studies. So, clicking from our interface while I magically launch this page and perform a search on PubMed using the gene and the syntax. You can further filter from there, you can read abstracts directly in our interface and a link out to the paper to read the full package.
So onto interpretations for this first TP53 variant are available on our system, while this is, again, a variant that does not have guidelines or drug labels. So you can learn about such variants from a paper cited here. Clinical trials are also available for this variant and can place it on level 2C on the classification scheme. And then I'd like to point out, as I mentioned before, a feature of our shared clinical data network is that variant information is shared in an aggregated fashion, de-identified, across all views of our platform, and so you can see for this variant, you can explore that frequency with which others have identified, across different disease types. Likewise, in the blue bar, you can see the frequency with which your lab has encountered this variant in a given disease type. And on the right, you can learn how others have deemed the clinical impact or relevance of these variants and how they've classified them across the classification scheme.
And if you look at clinical evidence sources, you find that there's some conflicting evidence from submitters, but the general consensus is that this is a benign variant. And indeed, the population frequency data we've loaded supports that conclusion, that exact NHLBI, thousand genomes not shown, all state that this is a very common variant. And this would be classified then as tier four.
So in conclusion from this example, the variants we identified here deemed worth reviewing are the fusion, classified as tier one with level A evidence, the other ALK variant, the escaped mutant, and TP53 G154V are deemed tier 2 with level C evidence, based primarily on availability of treatment options through trials. And then the P72R, the tier four variant, not reported. Here's an example for how we can represent these variants and create a draft of a final report for you to issue to the patient and to the treating physicians. We include information about exactly what was tested, some background clinical information, anything else relevant from the clinical history. We summarize the results, how many variants were called across the tiers, we print the basics of the evidence for each of the tier one and two variants, what is the evidence for strong clinical significance, the responsiveness and non-responsiveness to given drugs, or for the potentially clinically significant variants, what trials place them in that level.
And from there, the details about each variant in the finally drafted classifications for the three variants deemed relevant here, we show variants among certain significance and include details like the reference transcript, genomic coordinates if desired, and allele fraction if desired. And then we always call out what genome build was used and what databases were used to perform automated classification. And then any regions as the guidelines call out that are not covered sufficiently can be called out in the report. In this example, several exons that failed to achieve a coverage threshold that was set by the customer.
And then I'll finally point out that we support technical integration with LIMS systems and EMRs. This is both outbound messages or case ordering that come from your LIMS system to our PierianDx CGW, and then inbound back to your systems for delivering the report. This can be a rendered report, like a PDF for instance, or as the guidelines point out, it's important also to have discrete data on these variants. And so we support via our APIs or application programming interfaces, pulling of the discrete variant data and their annotations for data warehousing. And this is a really critical point as you gather data across many patients over time, your institution can gather. these valuable data and make discoveries, and have general institutional learning from evidence across many sequenced data sets.