WEBINAR TRANSCRIPT
CANDACE: Hello, everyone, and welcome to today’s webinar, where we’ll be discussing contemporary tools and methods for interpreting splice variants in hereditary cancer. My name is Candace Chapman, and I’ll be your host. Let’s get started! As most of you know, Mastermind is the most comprehensive source of genomic evidence, and can be used to quickly identify and review papers for patient diagnosis and treatment decisions.
In today’s webinar, our guest speaker from Myriad genetics will discuss the multi-faceted approach that brings together RNA analysis, functional evidence in the literature, published clinical cases, and other methods to classify splice variants. She will share how to facilitate the clustering of publications for variants that share a common mechanism, and how a critical review of all available evidence is crucial for interpreting splice variants. We’re really excited to hear this presentation, but first, here are just a few quick housekeeping notes.
Today’s presentation will include some Professional Edition features of Mastermind, which are necessary for clinical variant interpretation and reporting. If you don’t already have a Mastermind account, you can create one today by using the bit.ly link you see on the screen. That link will also be dropped in the chat window. This will start you with a free trial of Mastermind Pro, so definitely take advantage of that. If you’re joining us live, feel free to drop your questions into the Q&A as they occur to you, and we’ll get to those toward the end of the talk today. If we run out of time for all the questions, we’ll follow up with you after the event. Today’s webinar is being recorded, and a link will be emailed to you to review or share after we wrap.
Now, I have the pleasure of introducing our speaker, Dr. Randi Rawson from Myriad Genetics! Hi, Randi. Randi is a clinical genomic scientist who works on variant classification and reclassification for Myriad’s oncology products alongside a diverse team of scientists, genetic counselors, and lab directors. Prior to joining Myriad, Randi completed her PhD in postdoctoral training at the University of Utah.
We’re also joined by Genomenon’s field application scientist and Mastermind expert, Denice Belandres. Hi, Denice. You may be familiar with Denice because she leads our Mastermind support and training programs. Denice is going to kick us off with a quick overview of Mastermind, but before that, I just wanted to point out that Denice has just published an excellent blog on the best strategies for searching SNVs and ways to get optimal results. You can find the blog on our website at genomenon.com/blog after the presentation. Denice, take it away!
DENICE: Fantastic, thank you so much, Candace! Excited to be here. I’m going to show a few slides so that everyone can get a quick intro of how Mastermind works, what the user interface looks like, and understand how Mastermind is displaying non-coding variants like splice variants, which is, of course, our topic today. Here on this slide, you can see all the kinds of genomic concepts that can be searched in Mastermind. You can think of Mastermind as being an associations engine just as much as it is a search engine for literature. We take in information like genes and variants, then that can be linked across to other concepts, like phenotypes and therapies.
There are a number of ways to filter the results that are returned to you to help you prioritize the articles that are going to be relevant for you. For example, you might want to see functional studies or case reports. There are filters to help prioritize those types of papers. In that way, Mastermind really specializes in finding and connecting genomic concepts. So here, we have a screenshot of the Mastermind user interface, or UI, after launching a search for a gene and a variant. I have a splice variant here in my search bar. Randi is going to show a few screenshots herself, but I want to cover this broadly so we can follow along when she does present them.
At the top of the screen, you’ll find our integration with ClinVar, as well as information about related variants. Below that, there are four panes where we display information on variants and articles. Here, we’re zoomed in on the left side of this screen, which is focused on variants. All the published variants are displayed in a Manhattan plot, so that’s that variant diagram here. Below that, those variants are shown in a list or table format. Since I searched a splice variant in this example, you can see it highlighted in blue, and then also boxed in green on this slide. Notice how it’s designated in this table as N2543sa. SA stands for “splice acceptor,” and this is the splice acceptor site adjacent to the amino acid at residue 2543. Mastermind is treating that splice acceptor site as a group of variants that occur at the -1 or -2 position relative to that exon. It’s not just one specific nucleotide change.
These are the different groups of non-coding variants that Mastermind will bucket together. When we say bucket, we mean, for example, if you search for a variant like c.123-1G>A, that would fall on the SA, or splice acceptor, bucket for that search. You would get articles returned to you about that exact -1G>A change, but also articles mentioning changes at that same position, like -1G>C, and also changes at the -2 position, because those are the two positions that make up that splice acceptor site in the SA bucket. The idea being, if you’re looking at a variant in the splice acceptor, and maybe there are no articles about that exact variant, you may also be interested in seeing articles about other variants that might affect splicing in a similar way.
Since there are multiple variants that might fall into one of these buckets, you need to know when an article contains the exact change that you searched for, versus some other variant that’s in that bucket. We denote these articles by adding that crosshair symbol that you can see in the green box next to the article within the articles list. When you click on an article with that symbol, you can expect to find the exact nucleotide change you searched for in that article, and you can look at the sentence fragments to see that. Specifically, in the column where it says matched, you’re going to see the nomenclature that was actually used in that paper. That’s what’s shown there, below the articles list in the full text matches section.
Articles that have the crosshair symbol are prioritized, and they appear toward the top or first on that article’s list. When we scroll down toward the bottom of that list, you’ll eventually see some articles without that crosshair symbol. Here, I’m all the way down at the bottom of the articles list. You can see those two papers at the top, they have that crosshair symbol, but right below that in the red box is empty space. Those articles don’t have that symbol. When you click on one of those articles, you won’t expect to find an exact nucleotide match mentioned in that paper, but you can instead find a mention of some variant in that bucket.
Right now, we’re talking about the SA bucket. At the bottom under the full text matches section is where you will find that we found a match for a different variant at the same position. We had originally searched -2A>C, but here we have a match for -2A>G, so that’s something to keep in mind: the presence or absence of that crosshair symbol will tell you whether the exact variant is found in that article. With that, I will hand it off to Randi for her presentation!
RANDY: Okay! Thank you so much to Denice, Candace, and Kate for the invitation for this webinar. I really appreciate it! I’m really excited to talk to you all today about interpreting splice variants. As was mentioned, I work at Myriad genetics, and I work on the oncology team. The entirety of the talk is very much from an oncology perspective.
At Myriad genetics, we’re committed to helping patients understand their risk for hereditary cancer. With the MyRisk hereditary cancer test, we can evaluate 48 different genes that have been associated with 11 different cancer sites. This allows patients to understand what kind of health risks they have. If you look at the wheel on the left, this is outlining the 11 different cancer sites. You can see some of them are renal, breast, gastric, for example. The number on the outside of the circle is referring to the general population risk for each of those cancer types, whereas the number on the inside of the circle is referencing the lifetime risk for an individual who has a mutation in a gene that has increased risk from one of these cancer sites.
For example, patients with a mutation in one of our genes associated with renal cancer have up to a 70% lifetime risk of developing renal cancer, compared to only 1.7% in the general population. In the case of breast cancer, it’s up to 87% if you have a mutation, compared to 12.9% in the general population. As you can see, these have huge impacts on patients’ lives, and having this information helps patients mitigate their risks when possible. They can more proactively manage their medical care.
As you can imagine, because of the importance of this information and the kind of decisions that patients are making based on this information, it is really critical that variants identified in hereditary cancer genes are classified accurately. At Myriad, we are classifying variants based on modified ACMG guidelines, as well as available evidence. We use multiple lines of evidence for classification, and it’s really important that the data is critically assessed and integrated. In support of that effort, we rely on a diverse team of experts to classify variants. We’re relying on genetic counselors, variant specialists, PhD-level scientists, lab directors. This team of people is involved every single day in the assessment and classification of variants.
For today, I am going to focus on the classification of splicing variants in particular. Splicing variants make up about 10-15% of pathogenic variants, and they present some unique challenges to the classification process. To address those needs, we employ some specific tools to assess them, for example, relying on computational splice prediction and RNA analysis to help understand these variants. In addition, we look at literature in some ways that are specific for splicing variants. At a single junction, multiple variants may share a common mechanism, so it’s important to assess the literature for similar variants.
Luckily, Mastermind really makes that process seamless. To highlight some of these unique considerations and tools for classifying splicing variants, I’m going to go over three variant case studies. The first one will be an STK11 variant that highlights the power of Mastermind. The second one will be an APC variant, and the third, a BRCA2 variant. Both the second and third variants are great examples of the importance of properly assessing RNA data, and then the third is additionally an example of using clinical data to classify a splicing variant.
Let’s jump into the first variant case study, which is for STK11 c.597+1G>A. This is, again, a great example of the power of Mastermind for quickly gathering all the relevant literature for splice variants like these. To start out, just a little bit about STK11. Of course, it’s a hereditary cancer risk gene. Individuals with STK11 have a condition called Peutz-Jeghers Syndrome (PJS). This condition is characterized by hamartomatous gastrointestinal polyps that have a distinctive Peutz-Jeghers histology. This is a unique disorder, and it can be identified by histology, so you can find patients in the literature that are specific to this PJS phenotype.
In addition to those polyps, especially in childhood, individuals with this condition have pigmented spots at various parts of their body, for example, mouth, eyes, etc. In addition to having PJS, if an individual has a mutation in STK11, we also know that they have increased risk of cancer across multiple sites, some of the most common being colorectal, pancreatic, breast, and gastric cancer.
Let’s get into this variant that we’re talking about today, c.597+1G>A. This variant is at the donor site of exon 4, so it’s abolishing the donor site. Exon skipping is the predicted outcome of having this variant. If exon 4 is skipped, it’s going to create a frame shift, because as you can see here, the flanking sequences on exon 3 and exon 5 are not compatible with each other. An extra base will be introduced, causing a frame shift and a premature truncation. If we look for this variant in Mastermind (here’s a screenshot of the Mastermind website) you can see I’ve searched for STK11 c.597+1G>A. We have our article section and our full text matches section. As Denice mentioned earlier, there is this bullseye target symbol, which is telling us that it’s an exact match for the variant I searched for.
This is an exact match for the +1G>A. As Denice mentioned, these are prioritized. They’re sitting at the top of the list. There’s 15 articles total that come up for this +1G>A variant, and the exact matches are at the top. I’ve clicked on this particular one, and you can see in the full text match section, it’s telling me exactly what variant Mastermind identified in the paper. I can verify, yep, that’s the variant I’m interested in. Then, I can also see that this is from a paper by Papp et al. 2010. We can quickly gather the literature that’s exactly matching to my variant of interest. In this case, the variant was found in a mother and daughter who both had PJS in the Papp et al. 2010 paper. We have one PJS paper for +1G>A, with some segregation in a mother and daughter.
One of the great things about Mastermind is this bucketing that Denice talked about. Not only is Mastermind showing me the papers that have the +1G>A, which was four in this particular case, but Mastermind is also returning to me all the finds for variants at the splice donor site, or SD. The splice donor site is going to include any change at +1 and +2. These changes are, for the most part, typically going to have the same mechanism as each other, so they’re always of interest. It’s so easy to gather all that critical information and help us properly assess these donor variants or acceptors, in other cases. Bucketing them all together, we have 15 total. Now, I’ve highlighted here, I’m on a paper, you’ll notice it no longer has that bullseye exact match. I already know when I click on it that this paper probably is not going to have my +1G>A. Instead, this one has a +2T>A in the paper.
I just want to expand this full text match section, because I want to show some of the advantages to this full text match section. This particular paper is Orellana 2013. If we zoom in on the full text match section, you can see again that the variant in the paper is the +2T>A. Here, Mastermind is showing us direct text fragments from the paper. This is literally what’s in the paper. I don’t even have to click and open it yet, I just have immediate access to some really quick information. I see the variant, and I can also quickly see, hey, this variant looks like it’s probably in a PJS patient or family in this paper. Just quickly clicking through the Mastermind website, I immediately know this Orellana 2013 paper is of very high interest to me, and I should prioritize it during my analysis process.
For the patient data for this variant, so far, we have our one find for the +1G>A, which was found in a mother and daughter with PJS. When we combine variants at the splice donor junction, we end up finding two additional patients. There is a PJS patient in that Orellana 2013 paper, and there’s a +1G>T in a child with PJS in Jiang et al. 2019 as well. Importantly, all three of these variants are predicted to have the same splice impact, so they’re all abolishing that donor site. None of them is creating a cryptic or doing something else. They all have the same splice prediction. This literature can be combined, giving us a total of three PJS finds, plus that little bit of segregation in the Papp paper as well. This was enough for us to upgrade the variant from our default suspected deleterious to deleterious.
For the next case study, we’ll look at this APC missense variant and discuss data from Myriad’s RNA lab. APC is associated with Familial Adenomatous Polyposis syndrome (FAP). Alternatively, it can also be found on patients with an attenuated phenotype, or Attenuated FAP (AFAP). Both of these conditions are characterized by large numbers of adenomatous polyps in the GI system, particularly in the colon, rectum, stomach, and small bowel. Individuals with mutations in APC also have increased cancer risk across multiple sites, but in particular, colorectal cancer.
This variant that we’re discussing today is super interesting because there’s actually not an obvious splice defect for this variant at first glance. It is a missense variant, c.1902T>G, and it’s not creating a cryptic site, it’s not impacting any of the native donors, etc., etc. It wasn’t going to be an obvious splice variant. It’s located in exon 14 of APC. But I’m about to show you that there is actually literature showing that this variant seems to be causing exon 14 skipping.
You can see that if exon 14 is skipped, exon 13 and 15 are spliced together. Those flanking sequences are not compatible with each other, and a frame shift would be created. In the literature — you can find this literature in Mastermind — it’s been observed that there’s loss of exon 14, in both patient RNA sample as well as in a minigene assay, in Grandval et al. 2014. Importantly, the paper didn’t cover enough information for us to know whether this was a complete or a partial splice defect. There was no allele-specific quantification. In other words, we don’t know whether the mutant allele could still produce some normal mRNA product. This could be a partial splicing defect. Since we don’t know the severity of the splicing defect, we classified this variant as uncertain, and requested an RNA sample.
We did get a sample, and we were able to analyze this variant, but before I jump into that data, first I just want to let you know what I mean by allele-specific quantification, and why it’s so important. If you look at this mock gel, we have a control sample and a patient sample. You can see that in both samples, we have an upper band that is representing a full-length, or “normal,” transcript, if you will, and then the patient sample has an additional smaller band. In this case, for example, that would represent exon 14 skipping. The thing we don’t know is whether the splicing defect is partial or complete. More importantly, what we really want to know is whether the mutant allele can produce any normal mRNA. We really want to know what’s going on here, what’s going on in the patient sample with the full-length transcript.
If the variant causes a complete splice defect, then we would expect to see only the wild type, or T allele, present in the full-length transcript. Alternatively, if the variant causes a partial splicing defect, then perhaps we might see something where most of the full-length transcript is coming from the wild type T allele, but there’s some very minor contribution of the mutant allele to that full-length transcript. Another option is that we could have a partial splice defect, where the mutant allele is actually contributing a significant portion of the full-length transcript.
Here, you can see it’s not quite half, but a large chunk of that full-length transcript is coming from the mutant allele, or the G allele. It’s really important to figure out which of these scenarios is going on, and we can’t tell which of these scenarios is happening just by looking at the gel or a chromatogram. We need to know this, because based on which of these scenarios is actually at play, it can result in very different clinical outcomes for a patient.
As I mentioned, we did get an RNA sample for this variant, and our internal RNA lab was able to analyze it. They did observe skipping of exon 14, exactly what Grandval et al. saw. You can see here, this is just a chromatogram from wild type transcript from the patient. You can see that there’s a very minor little blip representing the G allele. We are seeing some representation of the G allele in the wild type transcript, but importantly, our RNA lab was able to do allele-specific quantification. They did dilution-limited PCR, and they were able to sequence hundreds of individual alleles. Based on that data, we were able to determine that only 2% of the full-length transcript was produced by the mutant allele. In other words, this variant has a near complete splice defect.
Not only do we have the splice defects, but there are also some clinical cases in the literature. This variant has been found in a Swedish FAP patient in Rohlin et al. 2016. This particular family also actually had another uncertain APC in cis with them, c.1902T>G, but then there was an additional patient in Grandval et al. 2014 that was in the French FAP registry. Taking the near complete splice defects together with some clinical cases in the literature, we were able to upgrade the variant from uncertain to suspected deleterious.
For the last variant case study, we’ll talk about BRCA2 c.426-12_426 -8 del. This variant is a great example of why it is so important to assess the completeness of the splicing defect. Ultimately, we were actually able to classify this variant based on multiple pieces of strong clinical data rather than splice data. As most of you all know, BRCA2 is associated with hereditary breast and ovarian cancer (HBOC). In addition, biallelic loss of BRCA2 can sometimes cause Fanconi Anemia (FA). Biallelic loss can be lethal, but when not lethal, it generally causes Fanconi anemia. This is a disorder that’s typically diagnosed very early, so in childhood or adolescence, and it’s characterized by physical abnormalities: stunted growth, progressive bone marrow failure, and early onset cancers. As you can tell, it’s pretty distinctive. It’s identified early, it’s well characterized. FA will come back into the discussion in a few slides, so that’s why I bring it up here.
Of course, a mutation in BRCA2 is going to be associated with cancer risk across multiple sites, not just breast-ovarian. There’s also increased risk for pancreatic cancer, male breast cancer, and prostate cancer. This variant is at the acceptor site, and it’s deleting five bases within what’s called the pyrimidine tract. Here, you can see the acceptor site is labeled in green, and this variant is in the intron flanking exon 5. It’s deleting these five bases that are highlighted in magenta. The important part of this portion of the acceptor site is that it needs to be as pyrimidine rich as possible to help draw in the splice machinery.
Here, we’re deleting four pyrimidines. As a result, in the mutant sequence below, you can see that there’s quite a few more purines being pulled in close to the acceptor site. This kind of mutation is known to weaken acceptor sites. Sometimes it’s not enough to fully weaken the site, and it can still splice. Other times, this is really a severe blow to the acceptor site. Predictions can give you an indication of how severe this might be, but it really requires RNA analysis to truly know how severe of a defect a deletion like this might cause. The variant is in exon 5. If we were to skip exon 5, that would again create a frameshift. As you can see, the flanking sequences in exon 4 and exon 6 are not compatible with each other, and we would have a frame shift and a premature truncation.
There’s actually multiple studies showing a splice defect for this variant. Skipping of exon 5 is what has been repeatedly observed. In the interest of time, I won’t go into all the studies that are out there, but I will just highlight one study in particular from Zhang et al. 2009, where they did RNA analysis on a patient sample. They originally amplified across exons 3-7, and they observed two bands in the patient sample, and then the single full-length transcript in the control. Based on sequencing, they were able to determine that this lower band was coming from the loss of exon 5, which is shown here in their representative chromatogram.
Luckily, they didn’t stop there. They took the analysis one step further, and they did a nice experiment where they did another amplification, but this time, they spanned the amplification from exon 4 to exon 10. The reason why they did that is because this patient had a heterozygous variant located in exon 10. If they could capture exon 10, they were able to distinguish the two alleles from each other. Because this is an intronic variant, we become blind to which allele it is on once we’re looking at cDNA and mRNA product.
They bring it out to exon 10. Then, they cloned their RT-PCR products into a vector, and they were able to isolate 10 colonies that had exon 5 inclusion. In other words, they’re isolating 10 colonies that had normal transcripts. From those 10 colonies, six of them had the wild type allele and four of the colonies had the mutant allele. They’re observing that 40% of the normal transcript is being generated by the mutant allele. That’s a pretty massive contribution from this mutant allele, that normally, we would have thought might have a pretty severe splice defect. This data is indicating, no, the splice defect might not actually be that severe at all.
Because this partial splicing defect was observed, we classified the variant as uncertain. With time, we’ve been able to use clinical data to reclassify this variant. At Myriad, we used a tool called the history weighting algorithm, or HWA. This tool is basically analyzing personal and family history of breast and ovarian cancer, and then comparing our variant of interest against known pathogenic and known benign controls. This tool has been validated, so it’s been shown to be greater than 99.5% accurate. In the case of this particular variant, our tool is calling it benign.
You can see that data shown here. This red curve is representing our pathogenic controls, and the teal curve is representing our benign controls. Our variant today, the -12 to -8del, is represented by this line. You can see that the personal and family history of individuals with today’s variant basically looks exactly like those of other individuals, who have known benign variants in BRCA2.
With this, it’s very strong evidence that this variant is, in fact, benign, but we even had more clinical data than this. We’ve also seen the variant co-occur with other pathogenic variants. We have found this variant in trans, or on the opposite allele, of two pathogenic variants in two patients. If you remember everything I mentioned about FA earlier — childhood onset, pretty severe disorder — these patients where we found the in trans with pathogenic variants, those patients were coming to us at middle age at the time of testing. They had no features suggestive of Fanconi Anemia. This is really strong evidence that this variant is not pathogenic. If the variant were pathogenic, these individuals would have some indication of Fanconi Anemia by the time they were coming to us for testing. Instead, these patients were pretty healthy. If we combine our co-occurrence finds with our history weighting algorithm, we are able to reclassify this variant from uncertain to benign.
The overall important message with this variant is, I think it really highlights the importance of being cautious when interpreting splicing data. It’s important to note that most public splicing data does not include allele-specific quantification. That Zhang paper, it was really great that they did that further analysis. Unfortunately, so often when we’re looking at splicing variants in the literature, we don’t get that level of information. When it’s not there, it’s really important to be cautious, and consider whether there could be a partial or complete splicing defect.
Then, importantly, once a partial splicing defect has been identified, it really necessitates having clinical data to understand whether that partial defect is deleterious or not. Sometimes, partial defects can be insufficient to prevent cancer risk. Other partial splicing defects are very minor, and there’s no clinical impact to them. It really requires getting clinical data to answer the question, is this partial defect benign or deleterious?
With that, I’ll just summarize everything that I’ve talked about today so far. We’re using multiple tools and lines of evidence to interpret splicing variants, and we’re sometimes using these tools in unique ways, specifically for the context of a splice variant. For example, with literature, we really appreciate the benefits of Mastermind for clustering those splice variants together, making it so quick and easy and efficient for us to gather the important literature for these variants.
Then, again, just a note of caution when interpreting splicing data in the literature. Often, whether the defect is complete or partial is not assessed. Just take some of it with a grain of salt, and consider whether there should be concern that a defect could, in fact, be partial. For clinical data, we have a lot of options. There’s clinical cases, there’s our history weighting algorithm, co-occurrences, segregation, and all of this data is so important in assessing splice variants. As I mentioned just a minute ago, it’s really difficult to know whether a partial defect is problematic or not. We really are relying on that clinical data to point us in the right direction.
For computational splice predictions, this is a really useful screening tool. It’s identifying variants of interest. As part of that prediction process, it’s also really important to assess the impact on the transcript. Today, all of my examples were exons, where if the exon was skipped, you would get a frame shift and a premature truncation. Of course, for other exons, that might not necessarily be the case. Then, you have to do a deeper dive into the transcript and protein domains to understand, if I severely impair a splice site in this exon, is it still likely to be deleterious? That part’s always fun, but it is just a screening tool and a prediction process, and not a method of classification on its own.
Lastly and most importantly is RNA analysis. Myriad has an internal RNA lab. We always stress the importance of doing allele-specific quantification. We’ve kind of beaten that to death, the importance of understanding how much of the wild type transcript is being generated by the mutant allele. We’re taking all these different tools and combining them to lead to accurate variant classifications for splicing variants.
With that, I thank you all for coming and attending the webinar! A huge thank you to Candace, Denice, and Kate. I would say working with Mastermind during variant curation is wonderful, but it’s also just so wonderful every time I work with the people at Genomenon. It’s such a great team of people. I really appreciate everything that they do for us, and all the different tools and resources that they provide us. Thank you all so much. With that, I’ll hand it back to Candace for the Q&A section.
CANDACE: Wow, Randi, thank you! We do have an awesome team, and we love working with you. Thank you so much for the excellent presentation, and thank you to our audience members who have submitted questions. We have a lot of questions. We’ll try to get through as many as we can, but let’s welcome Randi and Denice back. Randi, you may want to take a drink of water. While you’re doing that, I’ll just ask Denice a really quick question, which is: does Mastermind group other types of variants other than splice/intronic?
DENICE: Yes, absolutely! In addition to the groups I showed at the beginning on that slide, so groups like splice acceptor and intronic variants, we also group together frameshift variants. It would look something like h46FS to designate a frame shift after histidine 46. Again, that refers to a group of variants, not just a singular change, so it might be a four base pair deletion that could lead to that h46FS, or it could be a one base pair duplication also leading to that h46 FS. Both of those variants would get grouped together.
CANDACE: Okay, awesome, thank you! All right, Randi: what in silico splice prediction tools do you use?
RANDI: We use an in-house developed tool, currently.
CANDACE: Okay. Let’s go on. Are you able to tell why the APC mutant allele was contributing 98% to the transcript compared to the normal?
RANDI: Okay, so it’s contributing 2% of the full-length transcript. It would, of course, be contributing 100% of the skipped product. Well, I shouldn’t say that, there could be normal skipping. I forget what those numbers actually are in that case, but it’s contributing the vast majority of the exon skipping product, and then only a very minor tiny portion of the full-length product. Basically, that means that, with that variant present, splicing is mostly happening incorrectly. Only on the very rare occasion is normal splicing happening. I hope that answers the question.
CANDACE: Okay! Well, Bruno, reach out if you need more. All right. Denice, this is someone who searched for a splice variant in Mastermind, and none of the articles have the crosshair symbol: am I doing something wrong?
DENICE: The first thing to note is that the crosshair symbol, bullseye, target, whatever we want to call it, the thing to denote those exact matches, that’s a Pro feature. It’s not available if you’re using the basic edition of Mastermind. That’s one possible explanation there.
The next thing to do would be to make sure that you’re searching for your variant using the cDNA change, so when you type in a change, like “c.123-1G>A,” when you type that into the search bar, you’re going to get a drop down right under that. It’s going to suggest that exact nomenclature, the C dot, but you’re also going to see the bucket as an option, so something like p.r41SA, maybe. Both of those are options to launch the search, but you want to make sure you’re launching it as the cDNA, because in order to get that crosshairs, you need to enter your variant as a specific nucleotide change, so not as the bucket.
Then, I guess, the last explanation would be that there really are no papers with an exact match for the change you’re looking for. Remember that it’s really easy to scan through those text matches to look at that, specifically, at the matched column, so you can rule that out pretty quickly.
RANDI: I was just gonna add — I feel like the vast majority of intronic variants do not have exact matches, so I wouldn’t feel alarmed by that. Like Denice mentioned, it requires a nucleotide change, so you can also search by genomic position and and get those crosshairs too.
CANDACE. Excellent, thank you! Randi, can dilution-limited PCR be performed for all splice site variants to validate? What other experiments do you suggest to confirm mosaicism of these variants?
RANDI: Sometimes, it can just be technically tricky to get the dilution just to the right spot. Some variants don’t require dilution-limited PCR and allele-specific quantification. When you have a variant that abolishes the splice site, you know it’s literally impossible for it to generate any sort of wild type transcript, so a +1 variant or a +2, etc., that it’s not even capable. It can’t somehow, sometimes work. The site is literally gone. Those are cases where allele-specific quantification is less critical.
For mosaicism, I’m not sure if this is what the person means, but if they’re asking if the variant itself is mosaic, identifying that would require doing fibroblast testing or something like that. You would have to test cells outside of the blood sample and see if you’re seeing the variant in any of those cells, some, not all, etc. It would require that kind of additional testing.
CANDACE: Okay, awesome. For Denice: apologies if I missed this, how often is Mastermind updated with new articles? Is it different for Pro and basic?
DENICE: Yeah, great question! we update our database on a weekly basis, so it’s very up-to-date. It’s the same for Pro and basic. You get access to the same information, whether you’re using Pro or basic.
CANDACE: Awesome. Randi, a question that could be pretty broad, but I’m sure you’ll be okay with it: what about the variants that have never been reported in the literature?
RANDI: Yeah, those ones are obviously difficult. Because we have our internal RNA labs, we’re looking at the splice prediction. We also have experts who are reviewing the sequence for all variants that have a suspicious impact on splicing. That expert review and splice prediction, we often are sending variants to the RNA lab for requesting samples, and we will investigate them ourselves if we get a sample from the patient. We do have that avenue for variants that aren’t in the literature.
We also have the clinical data option. Of course, with internal clinical data, that does require time. When we first see the variant, we’re not able to know what the clinical data is going to tell us later, but with time and more finds, we do keep an eye on that clinical data for genes, where it’s possible. It’s not possible with all genes, of course. We do everything we can, but yeah, it is hard, because most splice variants do not have literature.
CANDACE: Okay. Anything to add on that, Denice?
DENICE: Nope, not on my end. That was a great comprehensive answer.
CANDACE: Denice, Constantinos asked: do you accept variants using the standard HGVS nomenclature?
DENICE: Yeah, absolutely! We accept HGVS nomenclature in the search bar, but there are a lot of ways that we are actually matching the variants in the articles themselves. We know that a significant percentage of published variants are described in non-standardized ways, so Mastermind recognizes HGVS and others, including IVS nomenclature or one-letter amino acids. In the sentence fragments, you can see that in action. You may search something like r43x in the search bar. The matches that you will see in the articles will include, if authors are publishing that variant as a cDNA change, as an rsID, if they say ARG43Ter, and even some of the funky stuff like arrows and slashes, those kinds of things are picked up in the sentence fragments, as well.
CANDACE: Awesome. Randi, can you explain why or how your team modified the ACMG criteria? There’s a few additional notes: did your team take into account the new ClinGen splicing subgroup recommendation from this year, and how closely do you match those?
RANDI: We have not yet taken any action to incorporate that very recent update for the modified tiering. It would be a lengthy answer, but basically, we are taking the information that we feel has a high impact on classification. It’s kind of our own recipe, if you will. I’m trying to think of a good example, where we maybe differ — for example, the absence of a variant in the general population. We don’t give that a lot of weight to say that a variant could be pathogenic. We know that there are very rare benign variants, for example, so that’s an example of one where maybe we would differ a little bit.
The other thing where we get specific in-house rules would be, we have rules for how common a variant needs to be before we’ll downgrade it. That’s not any different from ACMG, it’s just that we’re applying our own little list of rules to that particular ACMG guideline. That’s what I mean by modified ACMG.
CANDACE: Awesome, just two more questions! I think we can get through them quickly. This one for Denice is one I actually know the answer to! Very proud. What does SD mean in your annotations?
DENICE: SD stands for splice donor. The image I showed in one of the first slides, in case anyone missed it, with the different groups, that’s available in our FAQs. If you’re in Mastermind, you can go up to the top right corner under “help.” In that drop down, you can get a link to our FAQs straight from the application. The FAQs are also on our website, so you can see that image and study all the different groups in the FAQs.
CANDACE: We also have lots of videos showing how to use Mastermind on our resources page. One more question for Randi: can other labs get access to your history weighting algorithm tool, and is that based on your own patient history data?
RANDI: It is based on our own patient history data, and no, it’s proprietary. Unfortunately, no.
CANDACE: All right! Wow, what a great group of questions, great presentation. Thank you once again to Randi, and to all of you for attending today! Here is one last look at that bit.ly link to create your free Mastermind account, and a reminder to read Denice’s blog on searching SNVs. Of course, feel free to reach out to us with your questions to hello@genomenon.com. Look out for the link to the recording of this webinar in your email inbox soon. Hope to see you all again at our next event, and have a wonderful rest of your day!
