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.
Andy: Today we will talk about a couple of sources from the CGW Knowledge base, practice guidelines, which we also refer to sometimes as treatment guidelines, and clinical trials. How we can view the content and use it in your report in CGW.
Andy: Here's our agenda for the day. It is a one hour session. I'd like to give a brief overview of the CGW collaborative Knowledge base, to put these sources we'll discuss in more detail today, into some context, and review some content that was covered by Will at our last webinar in October, I think, regarding automated variant classification. To make sure we have proper context for the rest of our discussion today, I'll talk about how our annotation works of these specific sources of practice guidelines, and clinical trials, so that you understand where we get this information and how we handle it. And, how it winds up getting into the Knowledge base. Then, we'll look at a few options of how our current clients are using these sources to populate their reports, on how they display this information, and then finally we'll do a bit of a live demo in the variant details page of CGW to look at how to search for trials and guidelines, and find the most appropriate content to include in your reports.
Andy: We'll start with a bit of background information. Many of you are familiar with the CGW Collaborative Knowledge base, and the sources that we annotate and pull in to the Knowledge base. The breadth of the knowledge base is represented by our variant details page, which we'll look at later, which many of you are familiar. We have a broad range of sources both publicly available to some extent, and some that we curate heavily that are proprietary sources. As well as user created content that's the shared interpretations that users of CGW write and save to the knowledge base.
Andy: The range of clinical evidences as showing in this slide, where at the bottom we have public sources and cyclical evidence, this includes genome builds, annotation releases for genes, transcript proteins within the genome builds, pathways, population frequency data bases and computational predictions.
Andy: All the way up to high levels of clinical evidence, above this dash line are the sources that have the most clinical evidence that really support definitive classifications that can be automated in CGW. Those include, at the top of the shared interpretations, that users write and save to the knowledge base, the practice guidelines and FDA drug label, which we're combining into the category of practice guidelines for our discussion today. As well as clinical evidence from sources like Clinvar and aggregated patient data from past patients cases analyzed in CGW.
Andy: Trials then falls beneath this line, we don't use those trial information that is ongoing clinical trials that are recruiting in order to classify variants for instance but there a source of information that your oncologist customers may be interested in seeing on their reports you return to them.
Andy: Alright, so, that's a bit of the background on the knowledge base and where those sources fall that we'll be focusing on today. The classification scheme for somatic cancer assays, any of you already in production will be very well familiar with this, it's a five level classification scheme, where levels one and two are actionable categories where there's some predictive prognostic or perhaps diagnostic information either in the malignancy of the patient exhibits or another malignancy that develops one and two respectively.
Andy: Level three is a variant that may not have that level of supporting evidence for action but has been described previously in cancer, for instance in COSMIC or in Clinvar or in some other disease.
Andy: I'll skip to level five where there's evidence that the variant is found commonly in the population and is not likely to be a disease causing variant or pathogenic. And level four is the catchall for the us's where there's no other information in any source in the knowledge base. And so as Will would have described to you in the webinar he gave, the negative sources and the knowledge base can't automatically classify variants into these categories based on where they're found in data bases and our own curated resources.
Andy: Just to review those sources of basic classification rules that most of us our somatic panel used, first is interpretations that were written either by your institution or other institutions and that been published to the knowledge base. Those take priority and precedence over other sources of classifications.
Andy: Second are practice guidelines/drug labels, these are the sources we'll be talking about today from NCCN, ASCO and FDA. These can classify their intake levels one and two. Where as interpretations can actually classify their intake to any level depending on how they are written and what level of classification is associated with the interpretation when it's saved.
Andy: Third as we described on the last slide our other data bases, population frequency data bases that could drive variance into level five or Clinvar, COSMIC which in some classification rules schemes could cause variance to be populated to level three. And then we can write other custom rules per your specifications to put certain variance that you feel should be level five into that category, etc.
Andy: Before I go further, are there any questions about that background information that anyone's got issue, it should be very familiar to those of you who are already using CGW on a daily or weekly basis.
Andy: Let's look, before I get in too much detail at some basics for practice guidelines. What are we talking about here, when we talk about practice guidelines?
Andy: So these are both the way we group them are both practice recommendations from the National Conference of Cancer Network or NCCN and from the American Society of Clinical Oncology or ASCO. As well as regulatory approvals. That is FDA drug labels. This is for the use or the avoidance of use of a drug so a known responsiveness or non-responsiveness. Could also be for prognostic information, less so from FDA but more from ASCO and NCCN, or even diagnostic information. So all of those [inaudible 00:09:10] impacts if you will, are present in our knowledge base.
Andy: The impact of these practice guidelines to your report and CGW for your patient cases are first that these guideline rules, that is logic written in CGW, as we described, can cause those variants to be automatically classified as level one or level two and whether they are one or two is dependent upon whether there is a disease match between the patient's disease and the disease with which the annotation was written, the disease level of which that rule or logic was associated. Also, depending on your desires for how you want your reports to appear, you have the option of having drug information or prognostic information automatically populate into your report at report generation, on which they can be edited and changed as you wish on a case by case basis. We'll see some examples of how this is done right now in CGW a little bit later.
Andy: And then we'll get to this in the live demo but there's some useful parameters on which you can search or filter within the variant details in CGW to uncover the most useful or interesting treatment or practice guidelines, those include the source, that is FDA and NCCN or ASCO, the disease with which the annotation is saved and the annotation level which I'll describe in a little bit, that's a annotation specificity level that we record as well as we do annotations.
Andy: And as an overview for trials, we are the current source for clinical trials right now is clinicaltrials.gov. We annotate oncology trials from that source. As I described, trials do not impact variant classification. However, they can populate the report automatically if you've chosen to have that in your reports, and so again we'll see some examples of how that can work.
Andy: Again as you are reviewing this content in our knowledge base, it's useful to filter clinical trials on disease, on phase, whether it's a phase one, two, three or four trial, or something in between those, annotation level again and geographic location.
Andy: Okay. How do we take these sources and annotate them to be discreetly represented in our knowledge base? So here's some basics that apply both to guidelines and to trials. These annotations are are performed based on searches of this content for set of genes and set of drug names. So we annotate now something on the order of 500 plus genes and oncology trials and treatment guidelines or rather practice guidelines based on the genes that CGW users are reporting on their reports. And so this happens in two different approaches. One is the initial annotation when we add for instance a new gene to the list of genes that we annotate. We do a search across that source. Identify all mentions of that gene or drug, if the drug we've added to our drug search list and perform annotation from there.
Andy: In addition, we are performing weekly annotation or rather weekly searches of changes to these sources. So NCCN puts out flash updates whenever any of their documents are updated or new documents are released. Likewise, we do searches of ASCO and FDA and clinical trials to find what has changed since the previous week. Alright, so we're keeping up with these on a weekly basis and moving those changes, additions, removals like trials that are no longer enrolling to or from our knowledge base.
Andy: One concept that we've introduced or expanded I guess, in the past year or so is the level at which we annotate findings or claims or trials in these sources. These are represented in the annotation level column in the varying details page, which we will see. The first two levels are variant level and gene level. Alright and you'll see I got my spectrum here specificity in this case it's that specificity of annotation that I referred to a while back. The most specific annotation approach is at the variant level where a specific variant syntax is referred to in a source, so K-RAV, D12D is explicitly mentioned in the trial or in the treatment guideline or FDA label. We'll give some other examples here as we go as well.
Andy: One step down from that is the gene level specificity and that is where there's no specific variant syntax mentioned but the gene name is given so there's no ambiguity there. And in some cases there will also be some more general information about the type of mutation so for instance a slip three internal ten in duplication, which can be a lot of different size insertions, so it doesn't have the specific syntax but it is relatively specific in terms of how it can annotate.
Andy: Mutations and axions nine, eleven, thirteen whatever those are, or [inaudible 00:15:47]a mutation in UGFR. We can figure out from that in fact based on NCCN guidelines if it's a trial we're annotating, what they really mean by that and so we can do some more specific annotation there so that we're not throwing a net over the whole gene any variant in the gene. And that we would label as a gene level specificity annotation.
Andy: There will be times when a trial purposes will be so non-specific as to say, any RET or RET mutation and all we have to go on there is assuming that it's a non-synonymous change in those genes. That's maybe the broadest, least specific annotation that we annotate with the gene level annotation.
Andy: Importantly, treatment guidelines loaded on the production today, are all at least two levels variant or gene. We do have less specific annotation levels. We recognized sometime ago that when we searched simply for gene names, for instance, across clinicaltrials.gov, we missed a lot of trials that had useful information that we could infer would be possible trials to consider in context of certain mutations. And so we began doing searches based on drug names that we can associate that are known to target specific mutation or specific genes. There are two additional less specific annotation levels that we annotate and ease are currently on production for clinical trials unlike guidelines which are limited to variant and gene levels.
Andy: Alright, so the drug level annotation means that the gene is not referenced by name. But there is a drug in that trial that is offered and we know, based on NCCN or other sources, that this drug targets a given gene. And so we can then make a claim about making association between a mutation in that gene and with this trial. And then it would show up as something for consideration to put on the report in variant details page under the clinical trials tab.
Andy: Now to further extend that, the pathway level of annotation is selected when a gene is not referenced by name, again, but there is a drug that is referenced and that drug targets a pathway of which a particular gene is a part. A good example here is an inhibitor MTORC could be annotated for mutations and PIK3CA because PIK3CA is in that same pathway and active mutation in that gene might be, a tumor with that mutation might be responsive to an inhibitor of MTORC, even though that drug is not act directly on the gene that is mutated.
Andy: The purpose here again is to be able to increase the amount of content that we can display that might be of interest but it's not a speci......
Andy: ... That we can display, that might be of interest, but it's not as specific by definition it my need some additional review, rather than automatic population in to a report for instance.
Andy: Okay so let's look at some specific examples here for treatment guidelines. First let me just give you a brief look at the sources for treatment guidelines, specifically kind of what the documents are that we start with to do our annotation. Here's the snapshot here of the variant details page where ... This is for NRAS Q61K variant where you have ... This is just page one, or part of page one of the returned results where you can see multiple resources here are returned, and again you're familiar that you can customize which columns are on here, but you can see the annotation level is shown, and the impact of the management ... Management impact or drug impact. This is from our test server, by the way so that's why have a drug level annotation that shows up, that's not loaded on [inaudible 00:20:08] for today.
Andy: The FDA drug label is this PDF document that we can search for and download. This is just a snippet of what's shown, but you can see this is regressionive, which is demonstrated to be a drug that can be useful for tumors with rasmutation, activity mutation. So we searched through this document, it's returned to us via an automated circuit with in our annotator to manual review of these in order to essentially turn the PDF document text into something, a discreet rule in our system, some logic than can fire.
Andy: Likewise, the NCCN treatment guidelines, or practice guidelines documents look this. They're a big long PDF with a section at the top with the specific algorithm that defines some useful information. Then a long discussion section looks like this, which from which we then can look into specific cited references and the draw inferences and make claims about responsiveness, non responsiveness to specific drugs, or prognostic information.
Andy: ASCO treatment guidelines, or practice guidelines come from basically from articles written in JCO. These are clinical opinions, or provisional clinical opinions, so we are again searching these documents as they come out, in order to pull out the useful information and create additional discrete logical rules that can fire and infer content in your reports or within the variant details page, at least for your review.
Andy: So for these practice guidelines, we have our list of genes, we have our list of drugs that are useful for annotation, and perform and annotated query of these sources in order to get hits returned that can be reviewed by our annotators. Again this could be maybe a brand new gene we've added, or we're looking across all the documents to find any mention of that gene. It could be a weekly update, or any updates that we are notified of from NCCN, but now we're gonna see what changes have been made to the documents, and within those changes, are any of these drugs, a reported gene mentioned.
Andy: So the context of that PDF document is read by our annotators, and they also look at referenced publications and read those, and then they create specific annotations based on those publications or that practice guideline document or drug label.
Andy: All right so here's an example of ... This is a really over simplified examples I'll show of how these rules are written. There's basically a condition, so this includes among other things, the diseases, the gene, the variant. It can also include the consequence, coordinates, genomic coordinates, type of variant if it's a copy number change, or a single nucleotide changes, these are possible conditions that can be annotated. And then the annotation content itself, again this is very over simplified, but you have a source, a drug, and impact, and additional information text, or reference claims, all sorts of things.
Andy: So here is an example, simple example of a BRAF, a V600E mutation, and malignant melanoma where FDA as well as ASCO and CCN surely are recommended or have shown approval for using Vemurafenib, the mutations, rather tumors with these mutations are responsive to that drug.
Andy: So this is the same example we just saw. This is an example of a variant level annotation, right? This is a very specific syntax. The document says this particular amino acid change is what confers responsiveness.
Andy: And again, there could be others. There could be other BRAF mutations that are likewise responsive, and those would be annotated as separate rules, but could fire very similarly.
Andy: As an example of the gene level annotation from a practice guideline. Here we have an ASXL1 annotation in acute myeloluekemia, and here we don't even have, we have no specific variant syntax. We have the consequence of non-synonymous, so any non-synonymous change in ASXL1 could fire this rule where it's showing the NCCN doesn't have any drug information, but this is associated with unfavorable prognosis in the context of that disease. So again, that is less specific in the sense that there's no particular mutation that is noted, so we have made an inference. A logical one that non-synonymous changes in the gene are those ... Those are the, that's the context in which we want this content to infer, and this to appear in the report by default, if you configure to report that.
Andy: And just to extend this a bit, here's an example of a variant level of annotation that has two sets of conditions. So we can write a number of logical rules that operate in this fashion where more than one variant must be present, or more than one condition must be me for that annotation to appear. In this case, we have in the context of CML, chronic myeloid leukemia. Two ABL1 mutations, one is a rearrangement with BCR fusion, and then another is the T315I variant. And so in combination, NCCN says that such tumors, or such leukemia's are responsive to Ponatinib.
Andy: All right. Those are some examples again of the variant and genomes. I'll show you some drug and pathway examples in the trials annotation. So do have any questions we want to pause for, Sara, that you see, or shall we plod ahead?
Speaker 1: No I think we're doing well.
Andy: Okay. For trials again, we are searching oncology trials on clinicaltrials.gov.
Andy: And so here's a trials tab for this same variant. Variant details page. And I'm showing here at set of ... A set of columns including our annotation level, some information about drugs, inclusion criteria, et cetera. Again, these are all configurable as you're aware.
Andy: So for trial annotations, we query again by our gene list or drug list. Trials that come back as hits here are screened by annotators, and then valid trials are annotated. Trials that are not enrolling currently are annotated but sort of kept out separately and only added to the load in production when they are in fact enrolled, enrollment begins. And it becomes an option for an oncologist potentially to put their patient on that trial.
Andy: So here's an example of a variant level annotation for a trial. We have glioblastoma, IDH1 mutation, R132H and here's a trial I found it, it's a Phase II/III trial in Phoenix, of a particular agent, obviously creatively shortened the trial name, but this is a very specific syntax annotated trial.
Andy: At the gene level we have an AML trial that says that basically, I think it's PDGFRA and might also have been a kit on mutations. Like it wasn't more specific than that in terms of any particular mutations in those genes. It's a Phase II trial for a Philadelphia chromosome negative AML's with this particular agent that I will not attempt to pronounce. And so this is an example of a gene level annotation.
Andy: And I've just shifted those up to the top to make some room for ourselves. A drug level annotation here from a trial I found for NRAS. This is for colorectal cancer, Regorafenib trial in Phase II, and so this was to I think identify a biomarker that predict responsiveness to Regorafenib, so this didn't actually talk about NRAS in the trial itself. It talked about the drug, but we note from our knowledge from the NCCN guidelines that NRAS mutations are responsive, can be responsive to Regorafenib, so this was added as a trial that could show up if NRAS on a variant is identified in [inaudible 00:29:45].
Andy: And then the final extension in there is the pathway level annotation, where, and this is really the same example that I gave. This is a trial for breast cancer, where adding Everolimus to a hormone therapy, and this is a drug that targets, this is an MTOR inhibitor, and so this could be annotated potentially for MTOR as a drug level annotation, but for PK3CA, which is in the pathway, but is not itself targeted by this drug, we've annotated this as a pathway level annotation.
Andy: So this is all there, again, for you to be able to have more content at your disposal, and then be able to tell at what specificity level was this content annotated. So that you can plan more scrutiny to those that are on a drug or pathway level because we had to make some inferences about what the trial really intends to target because they were not very clear about specific mutations or genes that they were looking for, the mutations in for enrollment. And perhaps they are not, but it is something that is made without this approach, we would be missing some content. But it is useful for reporting.
Andy: So questions now about the annotation approach. Anything about annotation levels that are unclear? You may have ... You may recall that we sent an email, maybe a couple of months ago indicating that we had rolled out the drug and pathway level annotations for trials, and so those are now available on production. We have not done so for treatment guidelines or practice guidelines to date. We need to develop a way in which we can keep the drug and pathway level annotations, the responsiveness annotations, et cetera off of the reports by default. We want to keep the default populating content to the most specifically annotated, the most secure content for the variant and gene level, and so that's where we have that for today.
Andy: We do intend then in the future for a drug and pathway level annotations for treatment guidelines or practice guidelines to not impact classification, not show up in the report by default, but be in the variant details page for your review and for you to be able to promote to the report should you choose to. Or to write about it in your interpretation.
Andy: Okay. So let's move on then to looking at some examples, they're taken some screens shots from reports that give you an idea of, if you're not familiar already with what our options are, or possible options are for showing some of this content on your reports.
Andy: So the result summary section of the report is intended to give a high level snapshot of the key findings in the patient's tumor, and this is one of our formats for display of this content where one can show, this essentially corresponds to this column level one variants, level two variants, or level two related findings, here showing responsiveness or non-responsiveness to particular agents. In this case, and another malignancy other than the one, in the long ended carcinoma that was exceptioned here for this patient. And then additionally, prognostic practice guidelines can be added to the result summary section.
Andy: Clinical trial opportunities can be indicated as well, so you could choose to include variants that may not have the actionable per se in this section.
Andy: And then for those of you who have used a report template that has this content ... These drug names or the disease name for responsive, or rather prognostic or diagnostic statements can be edited in line here manually. So you can make some changes there. Likewise, the statement about presence of the clinical trial opportunity.
Andy: And then in a newer version of this section of the report, this result summary section that maybe a couple of you are already familiar with, we've added the ability, by clicking on this plus icon and a given cell and a table to manually add a statement about responsiveness, prognosis, diagnosis, or even risk. And so should you find that you don't like where something is automatically populating in terms of which column, or you want to add another statement that ... From some other source you found, or that you weren't able to promote directly from the variant details page for the given variant, you can add that manually. So this adds some flexibility. You're interested in being able to do this, and you don't have that capability now, then do talk to us.
Andy: Now then moving down to what we've traditionally called our clinically relevant results section, right? Where in addition to a high level summary, we can add content like interpretations, and more prose that you might write to describe the meaning of this variant to this patient's disease. We have developed some tabular displays where, first for ... As treatment summary table we called it, we can show a list of drug names to which this tumor may be responsive or non-responsive, or in these cases of these two rows here, to which there may be a drug option that's available through a trial. We haven't made a claim about responsiveness or non-responsiveness so that you can actually toggle that icon to something else. But this can serve as a high level summary to the oncologist of the available drug options that they may want to consider. Again this is optional content that you can include in your report template if you have customers or your oncologist that you serve want you to provide this information on reports. Again some do not, and so we have our users really span the ... Run the gamut there of leaving out of all drug information and trial information, all the way to really wanting to maximize the content. So this one's been developed to show maximize the treatment, rather the drug and trial content.
Andy: And then down here, this table, again is on a variant specific basis so we've kind of created this set of tables for each variant that was used, you know worth talking about or actionable in some way, and one can add and remove trials. One can have certain rules to have trials populated only within their state, only within a certain phases, only for certain diseases, et cetera. And then of course, manually one can change what shows here on a case by case basis. So these are some of our newer options for how to display this content in report.
Andy: Now, some of you might be familiar with this simpler display of chronically relevant results where all variant content, be it interpretations or trials appears in a single table. One can simply write in information about an NCCN guideline that is relevant to the experience and make that part of their interpretation, there's another means by which you can appropriate this content. Albeit less automatically, unless this is a interpretation that you've written previously and it's saved, then and you want to automatically populate in the future for a particular disease or-
Andy: Automatically populated in the future for a particular disease or malignancy.
Andy: Here's another approach, and likewise trials can be included this sort of clinically relevant results summary table.
Andy: Okay so that is the extent of the slides I have for today. Since we have some time and do we have questions that we should cover before I move on? All I intend to do beyond this other than field any questions is to spend a little bit of time in the variant details page, reviewing some strategies for when you have 80 trials that show up in the variant details page, how to whittle those down in to those that might be the most appropriate for including in your reports.
Sarah: Yeah, let's go through the workflow in CGW and then we can come back to questions.
Andy: Okay, sounds good.
Andy: All right. Is this showing up okay?
Andy: Great. All right.
Andy: Okay so I'm on the guidelines page, practice guidelines for a BRAF 3600E, so if we take off all of our filters here. You know we have NCCN, FDA, we have 64 results here for treatment guidelines.
Andy: And again, because we're on test environment I do have a couple examples of drug level annotations, but these are not available on production at the moment. In order to avoid any non-specific content appearing into reports by default.
Andy: So an approach here for looking through guidelines and deciding what you want to include, again, depending on your reports, some of this content may already have auto-populated, but you may have remove some of it and want to start fresh, or you may have a report template that doesn't necessarily have this content by default and you want to review it yourself, and put it in to your interpretations. You can filter on source. Here we don't have ASCO as an example, but we have FDA.
Andy: Annotations. That gets us down to 12. And then the annotation level, is another useful filter so I would recommend starting at the variant level and finding the most specific practice guidelines that have been annotated. And working with those, you'll see that about half of the FDA annotations are at the variant level. Rather specific responsiveness statements for these agents, and you can always then click on any row of any of these tables, or most of these tables in the variant details page for most of the tabs and rather than having to add a hundred columns, just to see all the content, you can simply scroll through here and see what's been annotated. Not all the fields will be populated because they're not all necessarily relevant for the particular annotation. We have pulled several sentences from, this is either the actual NCCN practice guideline PDF document or from a paper that was cited in that document where you know it's particular information about what a given study showed and the supporting information here as to why this annotation has been made. Or why this claim has been made about responsiveness.
Andy: Then there's often going to be a, well there will be a link to the document itself, from NCCN, which is behind ... Sorry, this is FDA, pardon me. The NCCN documents themselves are I believe behind a table wall. So even though ... No, that's not correct?
Sarah: No, you can create a long in on your own and then ...
Andy: Okay, okay, great. Great. So you can access that directly if you want to read more, likewise with ASCO and FDA, all of them have a link out ... Which I'm not showing that column right now, to save on space, but that is something you can do to dig further.
Andy: If we look at the, I think it's the drug level, we have small number of annotations. I don't see, do we have anything at the variant level? Or rather, gene level? We do. So a couple of gene level annotations, less specific, may not cite the specific mutation, but something about BRAF mutations, or activating mutations. And then likewise down the line to drug and pathway.
Andy: Now for NCCN, you might want to go to the variant level. What do we have, 50 results still, so you might want to look at, perhaps you want to add drug impact and look specifically at ... non-responsive statements ... Drug impact statements. That's 19 of the 50 results. So there's additional levels of filtering you can do to whittle down into those most useful treatment guidelines. And there is some level of duplication here where there may be multiple significant patient outcomes. There may be multiple roads shown. So that's something to be aware of. You can get rid of that to some extent by doing some additional filtering.
Andy: In clinical trials, I find it useful to filter by phase, as I showed on the slide previously. So ideally you kind of identify a high phase trial that's already been through some vetting in human subjects. And identify Phase III or Phase II trials. Also the annotation level would come into play here, so those at the variant level are perhaps the most, you know I'm sure they're the most specific and perhaps the most relevant.
Andy: And then you may really want to go to level of ... Is this a trial in France? Maybe I'm not interested. And this is a free text search for the moments where maybe I want to see if there's anything here in the St. Louis area. It looks like there's a ... Some trial or other in the Missouri, where there's a site in Missouri. So that's one way to get down to a trial that might be more accessible to your oncologists patient if you can pin down most of the patient's are likely to be from your area.
Andy: I should note that this geographic restriction can be applied to a report template so that automatically you only get trials that are within a certain region, within certain states, et cetera. So that's an option.
Andy: So from here I might stick with the higher phase and move to gene level and see if I get any more hits that are within my region. I've only got one trial there, and so on.
Andy: You might also want to limit it to drug, or rather by disease. So you just make it malignant melanoma. By default, the inferred practice guidelines don't limit by disease. So you get any of your practice guideline that is relevant to that mutation, and then it'll be level one or two, depending on whether there's a disease match. But if you want to whittle down just to those ... Sorry.
Andy: That are relevant to malignant melanoma, then you can do so. It might require removing certain filters to find examples. Here we go. Phase I/II trials that are at the drug level. Just a couple of those, or certainly if I open this up now, I should get quite a number. Yeah. It looks like I do. 67.
Andy: Okay so those are a couple of strategies that I think our variant analyst team, we have a team of folks who are also part of our annotation team who can do review of reports, and preparation of interpretations and reclassification of variants. That's a service we offer. I might ask them for some info on how they go about identifying trials and guidelines through consideration and the [inaudible 00:47:27] strategies they use in order to find the best content. So you might ultimately vary, and may have your own approaches, and certainly interested in hearing what you found useful and solid approaches to identifying the best content. And over time, we can add more automatic logic that can use those strategies to really pull out the best stuff and put it to your report templates automatically.
Andy: So we have only eight or so minutes left in the hour. I think I'll stop here and open it up for questions. I guess I do want to make sure to mention the next webinar. Should I do that now?
Andy: All right. Let's see if I can get back to the ...
Andy: All right, our next release of the application CGW 6.0 is due out on test later this month, and then production in March. So in February, we'll be doing a webinar to demo the new features of CGW 6.0. Those include support for additional added support for constitutional exome, including automated, at least partially automated ACMG classifications.
Andy: Oh, well I put the wrong date on there. It's 2017. It is not in the past. So this is coming up here in about four weeks.
Andy: The ACMG classifications is a big feature of 6.0. We also are rolling out clinical data aggregation where are showing information in a de-identified fashion across all past cases run in CGW to give you some indication the frequency with which variants, you've met variant appears in certain tumor types. And then finally some additional features that we have incorporated after our merge with Tute Genomics that happened a few months ago, where were now incorporating some of their features, including their phenolyzer service, which is great for identifying likely pathogenic variants, based on a phenotype based workflow for constitutional exome, as well as being able to filter based on inheritance models. So those of you who are interested in constitutional will be happy to hear of these new features.
Andy: All right, with that, lets leave the last few minutes here for questions. Thanks very much.
Sarah: So the first question we have Andy is whether it is possible to see the level of annotation whether it's variant gene, drug, or pathway in CGW?
Andy: Mm-hmm (affirmative). Sure. So at the moment the place you see that is in the variant details page. Okay? So there's an annotation level column. Right now in production that column is not a default column and we're gonna make that change so it appears. It's showing here but edited it before our webinar.
Andy: So yes-
Sarah: Maybe we can demonstrate how columns are added?
Andy: Oh absolutely. Sure. So here under this customized columns, pull down, drop down menu, it will show you that if you have no text in the search box, all of the checked boxes are those columns that are already showing. And then from there, one can make it do a search and add additional columns. One can also simply browse through if one wishes, the available columns and check off any of interest.
Andy: And likewise, any of these columns that you can add here, can also be filtered on. Okay? So I think we've gotten rid of our phase filter here, but you can add phase, you can add any of these. If you have a trial ID that you want to go look for, you can do that too. Anyhow, the annotation is here, you can both filter and sort on it. Sorting will be alphabetical. Filtering will be a more specific about, you know you can check off the specific permissible value that we ... One of the four terms that are allowed in that field. And that's available now for clinical trial and for guidelines. The other sources, the annotation levels doesn't really apply.
Andy: Other questions here?
Sarah: Yes. There's one more question. It says if practice guidelines are not restricted to a disease, then how does the system decide to assign a variant or a classification level of one or two?
Andy: Mm-hmm (affirmative). Sure. So, the ... The practice guideline annotation does have an associated disease, right? In this case, we're in trials, but we're looking at most of these for BRAF, not surprisingly, are written on. Oh hang on, these are not in fact [inaudible 00:52:50] but there are 24 colorectal cancers you see here. And so essentially what CGW will do is try to match the disease of the patient to do the disease of the annotations. And so if this is a patient with malignant melanoma, and the only annotations here are for colorectal cancer, then those guideline annotations will drive the variant to level two, because it's not a disease match there. Of course here for BRAF we've got plenty, certainly, that are written for malignant melanoma, and so CGW will look at all of the evidence and pick the most severe classification. It will match the malignant melanoma disease from the patient to the annotations [inaudible 00:53:38] for the matched disease, and call it level one.
Andy: So I should be clear that all of the FDA, ASCO, and NCCN guidelines are annotated at level one. But their default is level one if there's a disease match and then if not, it becomes level two. Just like the other rule disease matching. For instance, interpretations. Where level one interpretation for lung cancer can show up in your colorectal cancer, in the report template but it will, because it's not a disease match, it will infer as level two variant.
Andy: Same story with your treatment guidelines. If there's a match, it'll be level one or two, depending on the disease match.
Sarah: Okay, I think were almost out of time here, Andy. Thank you. Thank you everyone for attending. Please stay tuned, you'll be receiving an email soon about the next webinar that we're having in about a months' time on February 9th. Please help provide feedback on these webinars, if they're useful to you and if there are other things that you're interested in hearing about from us. Let us know and we can incorporate some of those suggestions into our webinar content for future months.
Sarah: Thank you all and have a good rest of the week. (silence)