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.
Overview
The point of this webinar is going to be talking about, outside of your standards SNV/Indel type of review of five analysis components where we feel like we've seen these a lot of times where people need a little bit more help and a little bit more understanding about the process. It's going to vary from the type of setup that you have if you're using Illumina or Thermo Fisher or Archer. What you do for each of these is going to vary a bit, but we want to establish some basic for what we had expected to do and things that you need to look at, and why you even need to look at these to begin with.
Co-occurring Variants
We're going to start with co-occurring variants. It is probably the easiest thing to understand in terms of bioinformatics and QC. It’s not going to be that different from yours SNV/Indel process. Just to note though that a lot of the people we encounter, whether they have a physical system in place, something like Clinical Genomics Workspace (CGW) or not, the ability to quickly identify and filter out those co-occurring variants is a really significant thing. We wanted to talk about an example and a lot of these slides that we do today we're going to talk about non-small cell lung cancer.
This is one where you're going to see that pretty commonly, in fact, even when I was working on EGFR and PCR, we'd find that occasionally we'd get a patient that would have both the L858R, and the T790M variants in EGFR. The problem is, these have completely conflicting results. When we're looking at this, there's a question which is, as a clinician what does this mean for your patient, and as a pathologist do you have a duty to explain what is known about those conflicting results? We actually absolutely believe that you do. This is an example of a variant table that we have from CGW. You'll notice that in the first line, the T790M is specifically nonresponsive to Afatinib, Erlotinib and Gefitinib.
Then, when you get that EGFR, L858R in addition, it's actually responsive to those three. We found that clinicians, oncologists, and even pathologists when we're looking at this we'd have a question of, “Well, what do I tell my patients?” Or what do I tell my clients about this if they've got these conflicting results? If you'll notice in this table you could be lucky you could just have a system like ours where it would actually show that responsiveness, and would make it really easy for you to identify this is the maintenance that works fine in this case, in this exact example, but we really ask people to adjust their interpretations specific to that type of challenge.
If you always know that when you have variant A and variant B that they should be interpreted with some other clause, then we need you to address those as co-occurring variants with a different interpretation. Let me go into a bit more detail here. If you only have a VCF, and again this is for people who don't have a system like CGW that is doing a filtration, I find that the visualization of this could be a little bit challenging. Here are my two variants. Maybe they're right next to each other. Maybe you've sorted these according to chromosome 7. But really when I look at these, I need a system to call this out for me.
I find that whether you have CGW, whether you have any other system that you need a visualization that helps you identify those particular variants. Just a note, you do have lots of systems that do this. If not, you need to run an if-then statement query against all of your variants. You don't need them to all have the same approximate variant allele frequency. That kind of question comes up a lot. These examples can be really common. There's a couple that you might recognize. This EGFR is really common. I see a lot of people who have 2 KRAS variants, and I want to interpret them always together with different results.
But, there will be times when it is more difficult, there’s example of BRAF V600E and any TERT promoter, in thyroid cancer. We actually run rules for those, and can put those into a filtration buckets that you always see it. It doesn't just get missed by the eye, and it's not a specific variant table that's only a one to one. It could be multiples of combinations here. We have done one of these in CGW. We have actually come in here and created a filter that just looks for these two variants in combination, and we can actually have multiples of these that are a combination line. So EGFR, BRAF V600E, KRAS, etcetera.
You can actually have a whole bucket that is just co-occurring variant. What we do when we talk about this is specify the idea that there is typically one interpretation for your T790M and a whole separate one for your L858R. This is just an example of how we break down some of our interpretations. You'll see gene level information, variant level information, evidence and databases, and then drug and guideline itself. If you have both of these variants at the same time, reporting these out in isolation is not going to give you the information that says, “Hey, these have conflicting results.” What we do, what we recommend to people when they're doing this interpretation and they see what we would call a co-occurring variant is to have a line within your section that says, “This patient also harbors mutations for X, Y, Z.”
In this case, what does change specific to this L858R mutation? For this one, we actually have a lot more detail. This is a bit of a combination between the T790M and the L858R variants, and their information. As you get towards the bottom here, you'll see this red box that says, although we don't have an FDA approved therapy for it, there's a clear indication in clinical trials that, that alternative is valid for patients who have both of these. We actually think that this is trending towards becoming an FDA approved type of therapy.
This is the really important piece that you want to make sure you call out to your oncologist, to your clinicians to have an interpretation that includes a level of detail that explains these are conflicting results or because you have combination you need to take action X, Y, Z. There's a pretty long list of these. We've actually talked about publishing some of these. If you're interested in seeing a list of co-occurring variants, and being informed and updated as those come out, please put a comment in for Josh. He'll keep track of that for you. I want to use this example. This is a more detailed example. This is actually the NCCN guidelines section, and this is for that BRAF, TERT promoter combination. In this case it isn't a drug specific.
It is a mortality specific response. It actually says things like, “You have a specific prognostic implication if you have these two variants.” You can see some papers on that in here. We actually have an annotation team that spends all their time going through and reading these guidelines. You would see these guidelines in our annotation setup, in our guidelines set up. If you're not using a CGW system you do absolutely need some system to ensure that you are being made aware of anytime there's a guideline that is specific combination of variants, because again, we need you to know that it's important and that you need to call it out.
This BRAF and TERT one, it's a little more complicated to find. Just to know what our interpretation team does. We actually have over 1,000 rules that we run. One of them is if you see this BRAF V600E, for example, this TERT promoter, or we could look at an entire region that does any of these particular variants, then add interpretation X. What we're saying is, if you're not using CGW, or doing your own interpretations, then you need to have a system in place to show exactly that, that it's going to produce a new interpretation, and that you're going to change the way you interpret because you've seen both of them together.
Splice-site Variants
Co-occurring variants are probably the easiest of these complex variants that we're looking at today. When I go into a splice site variants next, and really focus on exon skipping for this particular MET. When we talk about this, what is exon skipping? It's this idea that normal slicing would let you read this as one, two, three, four, five. With exon skipping there was a variant that changed the way this was read. And so, Instead of one, two, three, four, five, this is one, two, four, five, and it just skips right over one of the exons. We're going to talk about this in terms of a MET 14 exon skip, because it has a well known drug associated with it, which is crizotinib.
What happens here is exon 14 is responsible for breaking down some of the receptors or MET. If you skip that exon then you have more receptors and crizotinib helps reduce … Acts against this particular receptors. There are NCCN guidelines for this, but really important is you need to know when you can identify those, and recognizing whether a MET exon 14 skip is real. Those variants that caused those can be typically found in four areas. I would recommend if you want to see this, we've put the publication where you can read some of these things, but there's really four areas including the five and three prime splice sites, branching site A and the poly terminating chat.
Basically, variants within those conserved regions tend to produce an exon 14 skip. What that means is when you are doing NGS analysis you need some real basic bioinformatics and QC things that you do. Step one, I would ask that you always check to make sure that your BED file covers splice areas, because your caller is going to look and say how far into the intronic region am I going to look to see what is going to be produced? Then you're going to need to look at your color. Your alignment, everyone is using very similar alignment tools. Depending on the setup that you're using, whether it's Illumina or Thermo Fisher or Archer, everyone's going to have a very different color setup.
Illumina and Thermo Fisher actually do specify some very specific pairing. They look for some of those common MET exon 14 skip. Archer doesn't necessarily require any kind of pairs. It just looks at all variants across splice sites. You'll end up reviewing those differently. When you're looking at this as a BAM file and you're asking, “Is it real?” You can see an example of a BAM file here. We're not actually going to show people how to read BAM files here. We're just going to talk about looking at them. Again, if you're interested in how to read BAM files and going into more detail on that, please submit a note to Josh or question to Josh. If you're one of our clients were actually happy to call you and talk about this if you need that kind of discussion.
Also, just important here to make sure that you look at this, the same way you look at any other variant. With splice sites you're going to find that directionality is really important. Depending on the system that you're using, you might actually have additional tools. With Archer, for example, we pulled this directly from a paper and it had really interesting view that actually just showed MET exon 12, exon 13, and skipping directly to the 15th. You can actually visualize this a little better depending on what type of visualization tool you're using. Always looking at the BAM file, making sure it's real. During your validation testing, you should have established a minimum amount of coverage and quality for each one of your various including splice variants, copy number variants, etcetera, fusions.
You're going to have a list for each client that will be different, that will say, “Here is the minimum coverage, here's what I'm expecting to see when I review this.” Again, what you're going to check is, does your BAM file cover it, is your splice caller able to find this particular variant? Is the BAM file matching what you're expecting? That can be a quick check, and then setting filters to make sure that those are working. Wanted a quick discussion here about the differences between these different callers and different tools that people are using. And so, a note, with Illumina they have very specific splice site variant calling software that they prescribe within their user guide.
With Archer and Thermo Fisher they talk about it the same way they talk about all their regular callers. You're actually just looking at splice site variant, except that they are RNA specific colors. You will also always be looking at the BAM through this, and then the output is always going to be in the VCF. You're going to typically get a separate VCF for some of your other splice site variants, the fusion variants.
It might be important when you get these to have a merging system that can merge different caller output files, even though they'll still be VCF, you might need to combine them, to review. Just a quick check. Again, the QC criteria for all of these should be the same. You should check that they meet the validated read depth. I need to check for isoforms. Is it giving you an unusual, even though it's the same slice variant, is it different reads in different directions? Is it an unusual example of this particular spice variant? Then, for those of you who are asking, does my assay cover it? Obviously I can't capture every single assay that covers it, but we did include some of the example assays that you're looking at.
In some of these cases, for example, in TruSight Oncology 500, we actually add a special tag to the system based on the filters set. It tries to make it easier to filter this downstream. We call that a tag within the VCF. We try to do that to make it easier for you to visualize whether a variant is real or significant. Let's move on. Just another quick example. If you're looking at an RNA BAM you can make this a little easier to visualize. Along the top you can see that I'm going straight from exon 13 to 15. If you don't have all of these tracks, you might end up reading it with this little 13 here, and this little 15 here. But it's really important that you find some quick way to visualize whether or not this is actually skipping when it reads.
Not every variant does it. Again, that's why we always recommend checking the BAM file for review prior to acknowledging if something is acceptable or not. Wanted to make sure that if you're just used to more hotpot tech checking, you got some examples of well known that exon 14 skip. I like this visualization because it really explains sort of where these variants hit across the entire transcript. If you're just doing a basic look up, at the very least, make sure that your system can identify … I forgot, small cell lung cancer. The system can identify these common variants. Because again, MET exon 14 skip had been seen in five to 7% of non-small cell lung cancer.
You need to make sure that you can identify at least these types. I want to note on interpretations here that the data is constantly changing. If you were to look at an old knowledge base, and if you weren't keeping your knowledge base up to date here you would see that in 2015 we actually still call this a level four variant. A lot of the NCCN guidelines and FDA guidelines had not been published at that point. If you look down at the bottom, you can actually see the highlighted NCCN guideline from 2018 that specifically talks about crizotinib therapy, and NCCN qualifies that as a two way ... Sorry. NCCN, by the way, it's the national Comprehensive Cancer Network.
It has a list of guidelines for how you deal with cancer, and how you should be dealing with treatments. I'd have to talk about FDA therapies. What we said here is, this specific treatment is going to qualify you for crizotinib therapy in this case. We want to make sure that when you keep your interpretation you work on keeping them up to date. In fact, this is the NCCN guideline itself that would specify for this emerging biomarkers. This is something that has very specific treatment guidelines. All of this is in the PowerPoint. You're welcome to review this or ask for any other citation.
Gene Fusions
All right. Let's look at gene fusions now. Within the context of gene fusion we really want to make sure we discuss a couple of really well known ones, and why we're talking about this.
This can be absolutely critical to your treatment. If you're giving any kind of solid tumor testing we really absolutely recommend that you're checking for fusion. Some notes here on things that you might have heard of in November of 2018 there was a new guideline about NTRK fusion and the larotrectinib treatment. You'll see that as locked though we have C5s on that. ROS1 fusion have some relation to crizotinib and you've all heard about the EML4-ALK variant, and how did that particular fusion could have specific treatment therapies as well. One of the things we'll know is that fusions have a lot of varieties, and so making sure that they're real is a little bit more challenging, and you'll see a lot of different types of fusion as we go through.
One of the challenges that that hit is a nomenclature challenge. What types of fusions are we getting? In CGW, we actually do standardize this. You'll notice that we're always having to have first of all in a VCF two rows to demonstrate directionality and to identify a breakpoint. You'll see in this case I've got breakpoints that match in both directions, but this is an Intron and Intron. We would actually call this a fusion inframe, because it's going to modify the MRNA in a multiple of three. We actually identify these, because it'll be very different from something that actually does a frame shift mutation, and even though they're the same pairs, this is an exon Intron, and is actually going to result in A type of frameshift.
I separate these so that you understand what type of fusion it is. It makes a difference when you look at the BAM files, but just to make sure that you have a standardization for this, because the syntax here, the nomenclature can be confusing for people. Making sure that across the board you're always calling them the same and you're going to just call them generally fusions, until you get to the time when it's reporting time. One other note here on this, what we call microhomology events. You'll notice that the breakpoints on these are not matching across both directions.
If you have a system that if calling this just as a lookup table, you're going to need some system to identify and to be able to quickly review that there is a matched pair here, and that they're close enough that they're actually the same fusion. Not everything is perfectly balanced. We actually do this within CGW, but I want to make sure that everybody who's looking at our system that they identify those components, specifically and especially if you're using Manta as your fusion caller. This is where we do not have time in the course of this webinar to go through every single caller and all of its challenges.
Although, it's happy to be a followup discussion with people, but there's a really important piece and that is to know your college challenges here. At the bottom of this page we have listed some recommended reading from your vendors where you can actually go in and see what all of their parameters are. But as an example, in a Thermo Fisher caller, it actually specifies it for fusions. It's common for you to get a single fusion that is identified as a true positive, and then to get a second fusion, a second pair that is actually the same thing as the original fusion. You end up with two lines. You'll be wondering, do you have two different fusions?
A lot of times they're actually the same, and there are some specific requirements in their system about how you should identify those. Number one, if you are calling fusion, review one of these helpful hint readers that we have, and make sure that you understand and has seen all of the quality parameters and requirements specific to your fusion. If you're a CGW client, call us up and ask us, “How do I know that this is working?” Or, have you reviewed your fusion caller in a while? Once you have that identified and you’ve set all of those pieces up you always going to want to check the BAM on this and make sure that it's good.
Some of our clients also run a BLAST algorithms in NCBI to confirm which fusion form this is, and we'll go through that in just a moment, how you can do that. Then make sure that your filter is set up. We have a note that we have some basic filters we've set up at CGW to make sure we're identifying whether it is a quality fusion, what is what we might call a low confidence fusion or something that we're not sure of, that requires more review. Also, that we have asked to simplify the syntax when it gets to the oncologist or it gets to the pathologist so that it is easy to identify, in terms of drugs and clinical trials and prognostics.
Just an example of that fusion caller challenge with Thermo Fisher, this is their Ion Torrent system. The caller will return a data points specific to the sensitivity section with a value of low, medium, or high. They actually say in their description that they expect the default to be set at medium. You shouldn't check anything with low because it's not within their standard quality parameters. Unless you did your validation to accept low, you should probably be setting it at the very basic, at medium. There are recommendations within those reader guides that we talked about, but to the point there's 29 fusion parameters that Thermo Fisher has asked you to review in this particular case.
It's really important for a caller, whether you're doing Thermo Fisher or Illumina or Archer, to identify what each of their requirements are for their callers, and make sure you check against those in your quality filters before you start spending time looking at variants that maybe aren't even worth your time. I did want to do a quick check of BAM files. Again, if you have questions about how to read BAM files, please ask for followup on that. In this case, I wanted to show you some examples. Sometimes they're really clear and sometimes they're not as clean and clear. Depending on how closely your zoomed, you might be able to see them really quickly.
We do recommend if you're looking at BAM files that you actually show your MET here. You tend to open up your BAM files, but just one view and there are some buttons within IGB for example, that will show the split view and make it easier for you to view some of your various inner fusions here. Anytime you're looking at a fusion, make sure you try to look at it in MET view so that it's easier to identify the variants. I'm just going to show you another example. This one is probably cleaner than most of the ones you would see. This makes it really easy to identify your results and make sure the quality and quantity is acceptable here.
If you have also decided to check the BLAST, just a quick note, if you haven't done this before, there is a way in IGV to essentially copy your read sequence, than to put it into NCBI, which is the free tool that you use to BLAST and check this against the all homo sapiens, for example, all humans. Then it will actually return the results that tells you what this isoform is in which version of the splice that you should be looking at. That's if you just want to make sure that you do have the right variant. This is one of the ways that you can do it. Not everybody does BLASTn. It is really going to be dependent on, again, your QC parameters and set up that you're using.
At the end of the day though, it's all about getting it into a filter so that you know what you need to look at. I'm actually showing some Illumina filters within PierianDx’s CGW right now. This is a bucket we have called low competence fusion. We have similar bucket for high competence fusion. What you'll see is our low confidence fusions have a separate setup. We actually have tags in a system for Illumina caller where they say this fusion is a high confidence fusion. If that is not true, then we want to know a minimum fusion score, and what called it before we even know to look at it. Or if it was called by splice caller, and it had this particular fusion score, here's what I want to look at it.
As opposed to the high confidence fusions, which might just be something out of Illumina that says, “Here is when I expect you to look at it.” In some cases we've already just put in those tags that say, “Here's when you should look at that.” In some cases you're going to have to set up large if then statements to say, “If I have a fusion and it is not high competence and the score is at least a 0.3 or higher, and it was called by this color, then show it to me and I want to review it.” Just make sure that you have some type of system to set up this filter to easily identify.
Again, we always find that it's hard for the human eye to quickly identify these things, and so you want some kind of bucket to look at to review and ensure that these are the ones that we're spending more time on, and checking to make sure they're worth evaluating. We've talked a little bit about the nomenclature, and I found that giving this particular nomenclature and actually listing to breakpoint is not that helpful to my oncologist unless it is a specific breakpoint that equals a specific drug therapy. Most people actually prefer to just list the nomenclature that it is a fusion transcript.
I know if you're a CGW user, we've actually done in 6.8, the latest version we've actually updated that so that instead of just reporting this entire nomenclature, you can say, “This is a fusion transcript.” You would still be able to identify all of the drugs that are associated and appropriate for this particular variant. Some information here about how you do your interpretations, and this actually applies to almost all interpretations, but I really wanted to call it out for this EML4-ALK fusion, as an example. The FDA has approval for quite a few specific drugs here, crizotinib, ceritinib, etcetera. I found that it's really important especially if you don't have a result table like this up top to be able to come into a system and make sure that in your interpretation section it clearly states that there is a therapy information.
We just give some examples though the bottom example doesn't have any therapy information. It could be because for this particular example this patient was not here for the same disease type. And so, it might be specific to non-small cell lung cancer that they said that there's no drug that's going to work for this particular patient in this disease type. We do absolutely encourage the input of therapy information directly into your interpretation statement and into your results summary, just to make it easier for your oncologist to quickly view and identify what they can do for this patient. To track on to other types of fusions, we want to make sure you as a client are able to call out.
Again, I'm giving non-small cell lung cancer samples here, but in this case NTRK fusions, this was announced in November of 2018, and this is tissue agnostic. If you've heard of Loxo, Vitrakvi or the generic name larotrectinib. This is anytime you have this particular fusion bio marker. It doesn't matter what your tissue type is. If it's breast cancer, if it's a variant, if it's non-small cell, it's solid tumor and you have an NTRK fusion, then this is an FDA approved drug for this particular tumor type. When you're doing your interpretations on here, it's important to be able to say, and this is a more recent one that says, “Yes, this is predictive or prognostic, no matter what type you have.” But in this case I found one and I'm specifying that, this is an FDA approved therapy for any kind of solid tumor.
Again, when we talk about why are fusion so important, and this is a really new and exciting drug that is specific to any kind of NTRK fusion need to make sure it's real, and you need to make sure you're reporting on this because it's a really new therapy that people … That’s a really amazing promise for patient. A note here. Note here, this is a ROS1 fusion. There's a lot of different types of ROS1 fusions, but these might actually be specific to your oncologist or to your clients based on progression of the tumor. If progression, due to development of resistance has occurred, then in ROS1 fusion Lorlatinib is recommended.
Again, we're just calling out the different types of fusion that are really important for you to be checking within your assay, and make sure that you're calling out because there are some therapeutic implications that are well known and established. This is a 2019 guideline that we've updated to make sure that you are always aware and giving your clients up to date piece of information. Just some example of fusion, again, making sure that you've checked to make sure that they're real, that you know your specific caller and that you incorporate those latest guidelines. We’re going to move on to copy number variants.
Copy Number Variants
I would say that most people who are using a copy number caller will see that, again, specific to the system you're using Illumina, Thermo, Archer. You're going to have a different callers and you're going to have different visualization tools. I will show you an example, an image here where looking at the BAM is for me pretty useless. For other people maybe they've looked at a BAM and they're looking at it a little differently, it might work, but the really important pieces making sure you've got a good visualization tool that you understand how your copy number caller is working. Is it going to be giving a relative copy number? Is it a min-max, is giving an absolute number. Then, making sure you know how to report against those.
As with co-occurring tables, you might need a rule that you're using and keeping track up for interpretations to say, “Here's when I would report on it.” We've given some common examples. Again, in non small cell lung cancer you might have also been asked about this PDL1 amplification, or Keytruda, CCND1 mix. And even in EGFR amplification can have a level two classifications for non-small cell lung cancer patients. Let's get into a view of that. The visualization tools first. When you look at your caller, again go into that system and see what your caller is doing. The most important when we're looking at that is, what does your system provide? Is it copy number? Is it full change? Is it a min-max value?
Then how are you going to look at that? Down here at the bottom, this is how you typically look at expression values. You would look at, for example, a dot plot that would show a gain or loss aneuploidy type of data. That's not how most of our copy number visualization tools work. If you're looking at the BAM you might not find that very useful either. Within each system please go and make sure you see what their copy number of visualization tool is, so the caller that you're working with. I've used this example from Archer over here on the right that is going to specify an EGFR amplification.
You can actually see all of the instances above a line, and this is going to give you a value of 5.12, which is a relative value, right against a scale here on the left, it's typically some kind of logarithmic value. It's important to know what kind of value you're getting, because as I get this, I might either get a copy number here or copy number full change. If you're looking at a full change value or that type 5.12 from the previous discussion, depending on your caller you might have come in here and just thought 1.5 meant that it was a 1.5 copy number change. It's actually a copy number of three. It's dependent on the caller, and the type of result that you're getting that determine what this algorithm is, what this value actually is.
Then you'll need to be able to come in here and identify, “Am I looking for a loss?” For example, and that's what I want to review? “Am I looking for a gain, anything with more than three?” Depending on your validation you might have said that you can't detect copy numbers with any great accuracy unless it has at least four. This is going to be dependent on your validation, as well as the caller. We put in a note that says you can actually select your own as well. I want to point this out that this can be end and or a line. I can have multiple tags here that say, if the copy number full change is between 1.5 and six, for example, that could be another filter that I'd setup. We would recommend that you do that?
Some examples of the things that we're discussing here, down at the bottom you're going to see groups interpretation rule. I've got number five listed here. This is something we all have heard about. This is part two with a copy number greater than three. In this example we're talking about trapping out therapy. But you need to make sure that you're copying number caller is identifying whether it's greater than three, and identifying the value correctly, even if you've got three, four, five or 20, that your interpretation role always triggers. What you don't want is to have 20 different interpretations or 17 interpretation for every hard two copy amplification from three to 20.
You don't want to have to create a lookup table that has one of those. Instead we run a rule that says, “If you've got a copy number greater than three or between three and eight, put this interpretation.” You can actually do it different levels and say between three and eight put interpretation X, between eight and 12 put interpretation Y. This is what our interpretation team does. We say if you're doing your own interpretation to make sure that you keep a list of these rules and ensure that you're checking anytime it hits in this pocket that you're going to go ahead and run that check.
Some other pieces that are important to note, and this is specific, if your mother like mine is watching a commercial and she sees this Keytruda commercial and she asks, “Hey, are you working with that Keytruda stuff?” “Yes.” This is available on the system. It's now pretty well known even amongst the general population because of the commercials. And so, Pembrolizumab is an FDA approved therapy for people with PDL1 expression level. We're going to bring up a point here when we talk about expression, but we did comment, they can call us out that there was an FDA guideline for expression level that says Keytruda or Pembrolizumab is approved in this case.
You'll also note that there's a subdivision here. Most of you as clients are getting information that says this is non-small cell lung cancer, but you might not get something as specific as squamous cell non-cell on cancer. This treatment listed here in this lighter peach color is specific to squamous cell non-cell on cancer regardless of your PDL-1 expression level. There may be points where you need to call out in the system that you don't have enough information about any kind of subdivision of this disease to tell your clients what to do, but you should absolutely call this out in case this particular drug therapy setup is available because your patient is this subtype of non-small cell lung cancer.
Then there's, we talked about it in the beginning but it's worth repeating. This section down here in blue talks about requesting confirmation because as we all know in genetics genotype does not equal phenotype, and looking at copy number values is not the same thing as looking at expression level. And so, if you are putting in a recommendation based on expression level, it is important down here somewhere in your system to call out to your oncologist that you determined this via copy number, and they should do some sort of orthogonal testing like immunohistochemistry testing for PDL-1 protein to measure expression and confirmed that amplification that you would have seen as copy number.
Always important to say sometimes we will look at something and say it’s the expression level has determined this drug, and we can't determine that the copy number actually matches the expression level in this paper. That's what our interpretation services team will do, and that’s what annotation guidelines teams do, is look through and say, “What does this actually mean?” It's really important to check that interpretation and call that out for your client. I did include a couple of other copy number variant interpretation that I thought were interesting that you might want to look at. Not everything is about a drug. Sometimes it's about a specific prognosis that you might see.
I wanted to give some examples here of poor prognosis for MYC amplification, and again listing those within the interpretation setup, and maybe even putting poor prognosis in your results summary. Then this is an EGFR, this is also a tier two level C evidence in the top amp ASCO guideline. This is again from our interpretation services team, and they're literally just coming in here and talking about what EGFR amplification has MET in some other trials in some other PubMed IDs. There's not an NCCN guideline here, not an FDA guideline in here, but there is some evidence that's worth noting and talking about in this case.
TMB and MSI
All right. I want to take a moment and talk about tumor mutational burden and microsatellite instability, and make sure I leave time for questions at the end here. You always hear us talk about this as TMB and MSI, and they're not variants, they're not going to be lines that show up in the VCF. Just some important notes here. They're going to be calculated by an algorithm that's going to be dependent on the assay that you're running, and what kind of tool Illumina or Thermo or Archer, what they're all using to calculate TMB and MSI. It really requires looking over a very large section of genetic material to calculate percentages and say, “This is MSI high.” If it has a high tumor mutational burden.
Along here on the right, you can actually see a view of mutational load by cancer type. So, cancer types that tend to be TMB high. We talk about this because there are drugs in clinical trials that are approved and applicable. They don't follow the typical classification rules, we could call them tier one because they do have FDA approved guidelines, but they don't really match things like pathogenic. Having a tumor mutational burden of 41.2 per megabase is not necessarily going to say, “Oh, it's pathogenic.” Because inherently you're talking about having a high tumor mutational burden, but they do have a candidacy for immunotherapy that's available. On the bottom I've given an older example of what we might put in for TMB and MSI, or maybe just cite a paper that has it, some detail.
This is kind of a teaser for somethings we're doing at CGW and we believe anybody doing these large cancer panels should be reviewing. We'll talk about why real quick. On the left you'll see an actual paper from nature genetics that is discussing very specifically in association with high TMB and improve survival. The prognosis for high TMB was significantly higher in this paper. That's the kind of detail that you want to be able to provide to your clinicians and your oncologist, even though it's a very generic statement, we're finding more and more this type of information that's available. So in a similar fashion on the right people with high microsatellite instability have actually been given FDA approved therapy for that Keytruda discussion.
Again, what we're trying to do is provide our patients and our clinicians and our oncologist with the most detail about what we know about this tumor based on the NGS testing and what they can do about it. Within that context, I want to show you some of our more recent interpretations in this example. I just want to make sure you know, this is a colon cancer example, and we've actually listed some detail about other therapies that are available specific to colorectal cancer, and having a high tumor mutational burden. We actually classified it as tier 2 C based on the evidence.
Given some detail, we'll give all the standard information that we've been given and others, but none of this is an NCCN guideline or an FDA guideline that I'm showing in this, but it will show some specific clinical trials, and some other pieces that are available. Similarly, we have some interpretations on MSI high, and you can see that specific to colorectal cancer. Again, you just saw that FDA approved, Pembrolizumab Keytruda that is available in this case. If you're not checking for microsatellite instability as high, not just looking at a specific genes but looking at an overall range, and treating this like a bio marker is important.
Within CDW, for those of you are CGW clients, I should note that we are extending our fax model, the way we analyze variants to include TMB and MSI and more detail in the future. This is part of our core roadmap to even expand how you look at this automatically. I did pull this from an example case that we had, where we pulled up selected clinical trials. I wanted to make sure that people are aware there's a lot of clinical trials available in the world right now for TMB and MSI high patients. You can actually incorporate this type of detail into your clinicians.
We found that this is something that people are really appreciating when they're looking at clinical trials and they're doing any tumor board to be able to look at not just an individual variant, that sort of a broad classification like MSI high and identify whether this colon cancer patient has a clinical trial available for that. Even if the only information you offer for someone is that they've got TMB and MSI data, that can still be clinically significant enough to proceed. I'm going to pause here because we've talked about a lot of detail and just recap what we mentioned for each of these. For co-occurring variants, the discussion that we had was less about looking at the BAM files, but more about making sure you up a filter to identify those variants.
Then in your interpretation you're calling them out as part of the same interpretation so that your oncologist and your clinicians can identify that maybe conflicting results might exist, but that there is at least one drug or there was a recommendation if you have both of these to take path X. And so, just making sure you have your filters, that you have established rules and that you interpret them, that you put the interpretations together and show the context of a variant altogether with all the various that are being seen. We talked about splice site variant and we were specific to exon skipping in this case. Again, if you're asking whether this MET exon 14 skipping is real, the way we do that is first by knowing how our callers work, so each individual color has its own platform, it's concerns.
Checking our BAM file pile up to make sure that it is real and that it is not a strange isoform, for example. Checking the bi directionality in our BAM files, and then using some basic filters if you don't have all of that experience. At least make sure you're checking the well known MET 14 splice site. Really, it's important to look at some of these exon skips and for example with your TSO 500, we tend to look at your, I think that's an EGFR and the MET 14, and the MET genes, specific for very, very well known drugs that are associated with exon skipping and splice site variant. We talked about gene fusion, and in those cases we talked about, for example, these end track fusions, the ROS1 fusions, EML4-ALK and how the drug therapies on those are very important.
It's important to make sure that you're looking at those in the guidelines, but that in this, in gene fusions more than anything, it's important to make sure you get your nomenclature right, and it's important to look at those pileups in the BAM and makes sure that you are identifying real fusion, and that if you've got a list of multiple fusions that it's not actually multiple fusions in different forms, that it's either one or they are separate fusions, and you need to report them out separately. It’s just really important to make sure you're looking at the BAM. Probably out of all of these, gene fusions are the ones where we find we spend the most time looking at them.
Because they can be the most complex, and you have to look at the MET. We did talk about Copy Number Variants, and you all are familiar with a lot of these as the PDL-1, Copy Number Variants, HER2 or ERBB2 copy number variants, how those amplifications will result in different therapies. Making sure that you check this important, obviously the visualization tool. It's important to make sure you're not just looking at BAMs if that's not working in your system, or that's not the caller that you have. Really using the visualization tool that come with the caller that you have can be hugely important.
Then finally we talked about tumor mutational burden, and making sure that you as you get into these bigger systems that you're able to analyze tumor mutational burden and microsatellite instability, and really look at those numbers and translate them across therapies, clinical trials and prognosis. This is a recap of what we did. I'm going to just take a moment and hand it back to Josh and remind you that there is a way to ask your questions. Josh, you want to just repeat how to ask questions at this point in time?
Yes, thank you Amber. A terrific webinar by the way. I know how much preparation you put into this, and you'll be happy to know you’ve already received some high praise from a number of folks who've sent some messages in. Great job in the webinar. To that point, yes, please ask your questions. There's should be a GoToWebinar panel or pane on your screen there where you can submit questions. A couple of questions came in early that I want to touch on. One is, will these slides be available, will the presentation be available, etcetera. The answer is yes to both. We are recording this. We will put a recording up on our website and we will make the slides available for download as well.
To the point on the screen that amber just moved to, there was a question earlier that came on about asking about your product, are there regulatory issues directly related to reporting drugs related to a mutation? The one thing I'll say is, for our particular clients and who we support, primarily our laboratories that are developing laboratory developed test, so they'll use our software, they'll use, in some cases, our validation services or perhaps are doing the validation themselves, and a number of other services that we provide. Or independent really of our services to take what would be your product or an assay from a vendor and then validate that clinically, within a laboratory developed test.
That's just the clarification there at the beginning. Perhaps I wasn't clear at the beginning as well. As were waiting there are a number of questions that come in. I also wanted to just recap that, we do offer a number of services. Another point of clarification I want to make is that we are assay or sequencing vendor agnostic. We do have close relationships with a number of vendors and our software, our knowledge base, the interpretation services that we provide. We support a number of different assays across all of these different vendors for our customers. All right. Let's go ahead and get to the questions that are piling up here.
Questions and Answers
Question: Actually, this is an interesting one, Amber, and particularly in your experience there was a question early on about, who usually reviews BAM files? Is this a role or a skill set that a technician would have, or in your experience, I mean, who is typically doing this type of work and what skill sets do you need?
Answer: That is an excellent question. I definitely reviewed some BAM files when I worked as a clinical laboratory scientist. But, I find that for BAM file review you really do need some specialization. There's no regulatory requirement for who is allowed to review BAM file. But, I highly recommend, for example, we have a team here at CGW for some brilliant PhDs that are really amazing at reviewing BAM. I find that a simple BAM you might have done by basic scientists like myself, clinical lab scientists, but as we get more complicated, you want to have a BAM review discussion with a higher level person. Maybe you've got a technical scientists, especially as you get into some of these more complex fusion.
There's again no requirement, but I'm not going to lie and I'm going to be really honest. A PhD does a lot better with some of the fusions than I do. I would absolutely recommend if you're interested, maybe we should do a followup webinar on BAM review for pathologists or geneticists for scientists, and even from bioinformaticist. I really find that most people who are reviewing it there will be a technical review from a low level person like me and then a ticket to someone who knows more detail, if there's any complications with the BAM. I hope that answers your question. I really feel like it's a two parts. Start with the low level scientists like me, have them do a review.
They can get the basics, and then take it up to a technical supervisor, if it is a complicated BAM, in any way. Then do group reviews. We did this in my lab as kind of like a little mini tumor boards, where we would have, “Here's an interesting case and let's regularly review these BAMs, so that we understand how to work with them in the future.”
That's good. To your point, if you're interested, we actually have a survey that will pop up once this webinar is done asking about additional topics. If you are interested in doing a deeper dive into the these type of technical issues or scientific issues, either send me an email with a transfer in the survey, put your answer in the chat there. If you're interested in that, the more interest we get, the more impetus we will be to establish a training program and a webinars around these really complex topics. All right. The next question here is, there's a comment actually first that it's easy to make notes on co-occurrence variants when you know how to tell me about the clinical significance.
But, this is something you and I just talked about yesterday, or a couple of days ago, Amber. What if you don't have previous knowledge on the significance of this co-occurrence? Maybe you could just make a comment on that, and then also how do we handle it?
Obviously there's a line here that is you don't know what you don't know, and the knowledge bases are changing so quickly. What we don't have is any system that automatically looks at four variants together that have never been seen before, and automatically know that there is a call out for it. One of the things I was showing in the filtration section is the ability to dynamically add new variants to the list of things you'll look at in the future. While we can't predict everything that's going to be a co-occurrence variant, and have a significance together combined, you can say when you're doing that interpretation, as soon as you find something that lists as a co-occurring that you need to add that to the list of things you look at in the future.
We do that with rules with our interpretation team, and our clients when they're doing their own interpretations can just add that list of variants to the filtration bucket, the co-occurring filtration bucket. Unfortunately, it's not really predictive for all future cases, but it does a good job of making sure that the next time you see one that also has this weird notes that you call it out. Again, we try to give our clients some of these examples. We've talked about maybe even publishing a blog that has some of those, but predicting all future co-occurring variants, I don't see as a system that … I haven't seen a system that's really good at identifying all possible co-occurring variants in the future.
That sounds good. Actually, I just want to be conscious of the time here. It is at the top of the hour. We do have a number of questions. Maybe we can answer one or two more before signing off, if that's okay, Amber?
Question: The next question on the list would be … There's a question on tumor mutational burden, how often you update the TMB interpretation. They have entered the NCCN guidelines. I believe we showed this on the screen and maybe it wasn't that visible. Maybe you could just comment on how our interpretations were, how they often they get published, how often the knowledge base gets updated, and then what our clients are doing to update those.
Answer: For tumor mutational burden for MSI, right now the way those work in fact model is each climate interpreting that alone. We are extending the fact model specific to TMB and MSI. Meaning that, you'll be able to look it up the same way you look up BRAF V600E, essentially. Having said that in terms of our interpretation team and how we review it and our annotation team, we want to clarify the difference. The annotation team is updating NCCN, ASCO and FDA guidelines weekly.
We have literally dozens of people who sit in rooms and bless their absolute gold, read through all of these guidelines, 234 page papers and make sure that our system has the data from those guidelines in structured format for that can automatically show there is a guideline associated to this theory. That's updated weekly. If you're using our interpretation services team, you can actually see that variant example that I showed earlier for MSI high for example. In this case this one was done maybe two to three weeks ago, and so it will always be the most updated. I know this is going to be visible in the system. You can actually query against some pieces like this.
This is part of the fact model extension in that you'll be able to look at this interpretation the same way you'd look at a BRAF V600E interpretation. See that anyone has got an MSI high here or other things that we have for people with this type of variant, or this type of NGS change within non-small cell lung cancer, or colon cancer, whatever you have like that.
Sounds great. Thank you Amber.
Question: One last question just to end here. This probably goes back to some of your wet lab days. The question is, how do you, or how do our clients confirm reportable variant, mutations fusions copied to them etcetera, by gold standard techniques? If it’d Sanger, FISH, MLP, etcetera. Do you have any insight into how often or how they're doing that today?
Answer: It's actually kind of a controversial question in the somatic world. In germline people tend to do Sanger confirmation testing, and that is sort of still a gold standard of acceptable orthogonal testing for variants that you find within sematic, especially when you're talking about variant allele frequencies and tumor percentages that might be on the lower side. Sanger is essentially a 2X, three, maybe a 4X, if you've got multiple primers. Whereas NGS is running 30X, 100X, 200X, 250X, 1,000X. My experience in the community is that people are starting to believe, especially for sematic more in NGS and that the requirements, the technical requirements from CAP, CLIA, COLA, whoever you're using as you're certifying body, don't actually specify that you have to have an orthogonal test after validation for sematic.
Again, if you're doing something like a copy number variants, expression doesn't equal copy number variant, and so an orthogonal tests might be really critical. Similarly, for Fusions, if you've got a questionable fusion. I’ve seen a lot of clients who on this unusual variant, sort of a base level where they say, “If it doesn't meet these 2C criteria, or if it is any way unusual, we will follow up with a Sanger or followup with FISH or followup with IHD. It really just depends on the type of variant for these fusions. I think it's very common for people to do a FISH confirmation if it's confusing. For a splice site variants, it might be a really easy thing or confirmation, but not necessary.
If it’s co-occurring variants you could detect with a PCR, but I don't find anybody double checking co-occurring variants with orthogonal testing when I'm out in the world at any of my labs. In fact I don't talk to many people who double check all of those, especially when we get into these larger cancer panels, because you're talking about now going into four or five different tests just to confirm something that is well past a validated excepted depth, if that makes sense. Not require, but I do find people will do it for questionable results.
