Myriad Oncology Live episodes are recordings of an open-forum webinar hosted by Dr. Thomas Slavin. The opinions and views expressed in this recording do not necessarily represent those of Myriad Genetics or its affiliates. To participate in a future recording, visit myriad-oncology.com/myriad-oncology-live for a list of dates, times, and subjects.
Myriad Oncology Live episodes are recordings of an open-forum webinar hosted by Dr. Thomas Slavin. The opinions and views expressed in this recording do not necessarily represent those of Myriad Genetics or its affiliates. To participate in a future recording, visit myriad-oncology.com/myriad-oncology-live for a list of dates, times, and subjects.
[music]
0:00:13.3 Thomas Slavin: Welcome. This episode of Inside the Genome is a recent recording of Myriad Oncology Live, a webinar hosted by me, Dr. Thomas Slavin, chief medical officer for Myriad Genetics. The opinions and views expressed in this recording did not necessarily represent those of Myriad Genetics or its affiliates. To participate in a future recording, please visit Myriad Live for a list of dates, times, and subjects. I look forward to exploring the world of genetics with you all.
0:00:40.3 TS: Welcome to Myriad Oncology Live. Thanks for coming on. A little house...
0:00:46.0 Holly Pederson: Thank you so much for having me.
0:00:47.1 TS: Oh, yeah, yeah, yeah. No, excited to have you. So we have... Today we have special guest, Dr. Holly Pederson on. I will have Dr. Pederson introduce yourself in one minute. I'm going to do a little housekeeping as always to start. And so today, we're going to be talking about translational, hopefully everybody can see my screen. I'm actually in the office today, so a different monitor's set up, but can everyone see the agenda screen?
0:01:17.5 S?: Yeah.
0:01:19.5 TS: I'll take that as a yes.
0:01:20.4 S?: It looks like we're seeing the Myriad Oncology, but it's...
0:01:25.7 TS: Yes. Yeah, so great. Okay, thanks, Brad. So yes, today, we're going to talk about translation of PRS into clinical care. This has kind of been an evolving series. We did one... We did two previous PRS talks this year. The last one was with Alicia, who really did a deep dive in the science. After talking with folks, after that one, I think people like the science, but there's kind of a desire to just know like, "Okay, yeah. The science looks good, but what do I do with this? How is it being used?" These kind of more clinical questions. So that's really the focus today.
0:02:06.9 TS: We just posted out through the year. I'm going to put one more in November that we're working on. Right now, we're trying to... We're using more external guests, as people have probably noticed, which is also why I'm not doing them necessarily every single week. I'm trying to do two to three a month, is the goal. December is going to be just a short month with the holidays and everything, so we'll keep that one a little light, but this one will be good. We're going to have Gwen Turner on, and she is the head of Diversity, Equity and Inclusion at Myriad. Just a very interesting background, recently started, so she'll lead that one.
0:02:47.2 TS: And then we're going to get into variant classification, just because we haven't circled back to variant classifications for some time, and we'll talk about differences between AMP and ACMG. Erin Mundt, one of our variant-classifiers, will be on here for this one, and it will be good. She has some really interesting data and slides to walk through, and then we'll just open it up for questions. And today we have... Edie is on as well. And so... Edie Smith. I don't know if you want to unmute yourself and introduce yourself, but Edie will be facilitating some of the discussion today and then also running the chat, which is why I wanted to introduce her.
0:03:31.8 TS: So if this is your first time, feel free to unmute yourself, ask literally whatever question you want. We're here for you, so hopefully, questions get answered and you leave with more education and smarter about these kind of topics than you came on. And we try to keep these team-based. Today's is on polygenic risk scoring. However, if you have just a burning question about something else as it relates to hereditary genetics, feel free to... Or tumor genetics, you can ask that as well. So Edie will be running the chat. Edie, I don't know if you want to unmute yourself real quick and introduce yourself on what you do.
0:04:17.9 HP: You're on mute, Edie.
0:04:21.7 TS: Edie, you're on... It looks like you're on chronic mute. I will ask to unmute you. Did that work?
0:04:32.2 Edie Smith: Can you hear me now?
0:04:33.7 HP: Yes.
0:04:35.2 ES: Oh, wonderful, wonderful. Sorry, I had to use the combination of phone and computer this morning. Zoom is acting up a little bit, so thank you, TJ. My name is Edie Smith. I'm a medical science liaison for Myriad Genetics. I'm a nurse practitioner by training, 30 years of clinical practice before retiring about five years ago and joining Myriad. My focus is on... Primarily on the unaffected patients and cancer prevention and cancer detection, so I'm happy to moderate the chat, ask questions, I'll put them in the chat, and I'll keep an eye on everything.
0:05:12.8 TS: Yeah, great. Well, thank you so much. And before I forget, because I know I will and I always do. We do... I did want to mention these are recorded now, so if you... And I think it says that when you log on, but we do... We're putting all these up on our podcast, the Inside the Genome podcast, so anything that says Myriad Live in front of it is from one of these Myriad Lives. So these are usually 50 minutes or so. So at least they're all archived, and you can go back through if there was one you missed and was something you want to listen to. If you don't see Myriad Live in front of it, that's just a standard podcast, so that's usually me sitting down with someone for 15 or 20 minutes and kind of going through a specific topic or aspect of that individual's research and career, like we did one with Holly, I don't know, probably about a year ago or so. That's somewhere way down here. There's... Oh, yeah, Let's Talk Polygenic Risk Scores.
0:06:18.2 TS: So yeah, so we have tons of content you can listen to your heart's content. So without further ado, I wanted to introduce Holly. So Holly is amazing. She is at the Cleveland Clinic, so many know her as Dr. Pederson, takes care of many patients needing hereditary cancer expertise. I first met Holly as part of the City of Hope training course, and she's been a good friend since then. Holly, I don't know if you want to... Oh, you're already unmuted. If you want to tell folks a little bit more about what you do at Cleveland Clinic and your general interest?
0:06:53.3 HP: Sure, sure. So I'm the Director of Medical Breast Services at Cleveland Clinic. We basically do everything nonsurgical. It's sort of a unique model. We do diagnostics, personalized risk assessment and risk management, and I personally focus on hereditary risk and risk management. I did a clinical fellowship with Dr. Charis Eng here at the clinic, and then did the City of Hope course as well with TJ, and that's how we met. Edie and I also go back a long way, and Deneen's on, I don't know how many years she and I have known each other for. But I've had a longstanding relationship with Myriad, and I'm very excited about the new polygenic risk score.
0:07:45.2 HP: It was interesting, TJ, how you mentioned that there's the science and then there's the translation into the clinic, and I can't tell you how many hours Alicia and I spent... Alicia Hughes and I spent together trying to bridge that gap and sort of make that connection and interpret a lot of the science that's going on and make it clinically relevant. So I hope that we're able to kind of touch on those things today.
0:08:21.3 TS: Yeah, yeah, 'cause you know, it is a new aspect of the field, so with anything new comes growing pains and trying to understand it and how to utilize it. So I know you put together some slides 'cause we were talking ahead of time. Spoiler alert. So Holly has a slide deck. Yes, thank you for cueing that up. So we thought it was good, you know, Holly can go through some of the work that she's been presenting because it has a nice translational component to it, but yes, if people are on and want to ask questions, then for sure, at the end of this brief presentation, we'll pause, make sure that questions are answered. I have some other work that I've been putting together on thinking of ways to conceptualize all this and bring it into clinical practice, if we have time. So take it away, Holly. Thank you so much again.
0:09:25.7 HP: Oh, sure. I think two of the biggest challenges are really the education of primary care providers and other women's health providers caring for women in this space, and really working toward effective communication around risk. I just think that those are some of the areas where we're going to probably see some of the biggest advances. I just wanted to give a brief presentation discussing how single nucleotide polymorphisms or SNPs can add to the precision of breast cancer risk assessment, explain how polygenic risk scores are generated and validated, and outline the relevance of SNPs and the PRS to clinical practice with the potential to improve screening and prevention strategies.
0:10:22.1 HP: So in just a little over the course of my lifetime, we've really gone from genetics affecting very few people with rare disorders to it becoming everyday clinical practice and increasingly personalizing medicine, essentially. In 1953, Watson and Crick were credited with determining the double helical structure of DNA. In 1977, the laborious Sanger sequencing was introduced, looking gene-by-gene at changes, and through that process and looking at really high-risk families, BRCA1 and BRCA2 were identified in 1994 and 1995. I don't have it on this timeline, but in 1997 was when I started at the Cleveland Clinic with the high risk center, so that was a very opportune time to sort of get in on the beginning of this journey.
0:11:30.0 HP: Next generation sequencing and the completion of the Human Genome Project really changed genetics entirely with the ability to sequence multiple genes simultaneously at a much lower cost. And in 2013, multi-gene panels became available, which now are just commonplace, everyone uses multi-gene panels. But that wasn't the case prior to the fall of 2013, and now we're really looking at the next incredibly exciting addition, I believe, with the introduction of the polygenic risk score and sub-stratification of risk, both in gene carriers and in non-gene carriers.
0:12:21.0 HP: This is how we've always historically looked at breast cancer. There's this 10% to 15%-ish slice of the pie, which is attributable to hereditary cancer syndromes, true genetic mutations, which are felt to be causal and inherited from one generation to the next. There's another segment of familial clustering, which is not gene-related per se, but possibly due to shared environmental influences, gene-gene interactions, shared dietary and exposure patterns, but the majority of breast cancer is sporadic, that is, it's seen in women who have no identifiable risk factors, and I think that may be about to change.
0:13:19.8 HP: I think possibly, my hope with the polygenic risk score going forward is that we may be able to begin to identify low-risk individuals or lower risk individuals in the general population, and that may help with screening strategies. You know, I mean, when... You all know this, but when someone has, quote, genetic testing, what we're testing for are these highly penetrant and moderately penetrant genes, but there are over 300 single nucleotide polymorphisms that have been discovered to date that individually confer very small levels of risk, but in aggregate, may affect risk, both in gene carriers and non-gene carriers, and may explain that sort of missing heritability piece that we're really lacking.
0:14:23.5 HP: So these SNPs were discovered through genome-wide association studies, which is really a process where patients with a particular disorder are compared to patients without that disorder to see where there are genetic changes. And the SNPs that are associated with breast cancer were largely discovered in a very large population of Caucasian-European women, and that is not without its problems, as we'll discuss.
0:15:04.3 HP: So you can see BRCA1 by itself confers a very high level of risk, but you can also see that combining the effects of these SNPs also can be significant. And this was the first landmark study published by Mavaddat et al in 2015 showing that a 77 SNP polygenic risk score was really able to sub-stratify risk between the highest quintile of the polygenic risk score, resulting in higher lifetime risk estimates for breast cancer as compared to those at the lowest quintile, and this was validated both in women with and without family history, so this was exciting.
0:16:00.8 HP: But most of the data, as I mentioned, came from those of European ancestry, and so that was really an issue. As Edie had mentioned, primarily, the patients that we see are the unaffected patient with family history, and at most, 5% of those patients will have a genetic mutation. With the multi-gene panel tests, about 25% will have a variant of uncertain significance, and that may be higher in certain ethnic groups where we don't have as much data, and those clinically need to be treated as negative. And in those patients and the 70% that test negative, we resort to mathematical risk modeling, but that's really so... It is what it is, but it doesn't incorporate any of the patient's actual information, genomic information into that risk estimate, and that's what's so exciting about this polygenic risk score.
0:17:12.6 HP: So I was fortunate to present this data at ASCO this year on the sort of recalibration and validation of a polygenic risk score that is meant to account for the changes that we see in people that are not European. Because the polygenic risk score, as it was derived from European populations, is quite accurate in European populations, but the goal was really to develop a PRS that has a high level of accuracy for all women in terms of good risk discrimination, that is the ability to tell high-risk from low-risk individuals, and accurate calibration amongst women of different ancestries, because the allelic frequency of SNPs is different in women of different ancestries.
0:18:24.6 HP: And so you can't use a European-derived polygenic risk score in a person who's not European, of European descent. So 93 selected breast cancer SNPs were used in combination with 56 ancestry-specific SNPs in order to develop this model. And you can see that before the model was created, there was an over-estimation of risk in non-European ancestries. In fact, in women, in black women, you can see that they have an estimated risk nearly twice that of white women, whereas we know that black women have a similar rate of breast cancer development, if not slightly lower. And so this European-based polygenic risk score needed to be recalibrated to avoid over-estimating breast cancer risk in non-European women.
0:19:40.2 HP: And so this is where Alicia and I really got into the weeds and had some fun, but this is the... These are the results of the discrimination, again, before it was recalibrated. And you can see that the odds ratio per standard deviation in Caucasian women is the highest and is the most statistically significant, and this was improved upon by the new model, which really took women's ancestry and determined it genomically. Rather than using self-reported ancestry, it's felt that the major source ancestries for the contemporary US population really come from the components of DNA that are contributed from the continents of Africa, East Asia, and Europe, and any US woman can kind of be divided into her ancestral components from these three regions. And again, the allelic frequency of the SNPs differs in these different continents, and that needs to be taken into account in determining a polygenic risk score.
0:21:18.7 HP: And so each woman, rather than using self-reported ancestry, these ancestral SNPs were used to determine a woman's genomic ancestry, and then the proportion of each of these components was used in a weighted fashion appropriate for that continental ancestry to create a PRS for that woman that was more accurate than if it hadn't been recalibrated. So you can see that as intended, the new PRS was recalibrated so that there was not no longer an over-estimation of risk for non-European populations. There's a slight shift in the Hispanic population due to a protective Amerindian SNP, but if you take that away, it too centers around zero, and there was improved discrimination in all groups as well.
0:22:34.1 HP: It did not reach the same level of discrimination for all women. You see that with Caucasian women, there is still this odds ratio for standard deviation that is higher in most groups. We still don't have the discrimination that we would desire in black women, but it's a start. It's a framework that we can build on as more data becomes available. And part of it is the lack of data that we have, but part of it is just the genomic diversity of Africa itself, which is really an issue. We're going to need even more data in the black population to make things, to give the discrimination that we want in that population.
0:23:27.9 HP: So we developed a framework for a PRS that's accurate for women of all ancestries and can be adapted as additional data becomes available, and this clinically validated PRS provides calibrated genomic risk discrimination for all US women and possibly for women around the world. It also... The PRS may be an aid for decision-making in gene positive patients. And I have been taking care of gene positive patients, as I mentioned, since 1997, and they are so excited about the possibility of sub-stratifying their risk based on the polygenic risk score.
0:24:17.8 HP: And you can see from this paper by Gallagher last year that by using the polygenic risk score, we quote a woman with a BRCA mutation and estimated lifetime risk of 70%, but it could be as low as 50%, which is still very elevated, or as high as nearly 99%, and that may be useful for some women with BRCA mutations in decision-making. But where it really probably is going to be most useful is in women with genes like CHEK2, ATM, and PALB2, where there's an incredibly widespread from a low risk situation to a very high risk situation, and in non-carriers of genes, especially in patients with strong family history.
0:25:18.0 HP: This was such an exciting paper, and you can see, say, with CHEK2, we estimate that a woman has a 20% to 30% lifetime risk, but by using the PRS, we can further sub-stratify that, such that it might be 6% or 71%, and certainly, that can be helpful to people in terms of making decisions about their care. This comes from a more recent paper, looking at CHEK2 carriers, and again, about two-thirds do have that 20% to 30%, and even as high as 50% lifetime risk, but 13.4% are greater than 50%, and that's a threshold at which we would consider a discussion around surgical intervention. And 24% are low risk. And while those women still need to obviously be vigilant, they may make different choices, say, around preventive medication.
0:26:31.6 HP: It's also been shown that the PRS is more effective when combined with a family-based history... A history-based model, such as the Tyrer-Cuzick model, which is what I typically use in the clinic. And it makes sense; to use the women's own genomic information, not as an end-all be-all, but as a piece of risk information that can be added to the other risk factors, makes a lot of sense, and in fact has been shown to improve the discriminatory accuracy of the prediction. This is in using the Tyrer-Cuzick model alone and using the PRS alone, and this is two validation cohorts using both. And you really see the benefit of adding both.
0:27:26.5 HP: This is an abstract that Alicia and others have submitted to San Antonio, discussing the validation of that combination with the new polygenic risk score in combination with the Tyrer-Cuzick risk model. And where I think clinically we're going to see the most exciting initial use of the polygenic risk score is in helping women to decide about preventive medication. Chemoprevention uptake is simply miserable in all groups and women are afraid to take it. They're afraid of the side effects.
0:28:09.8 HP: This was a study out of the Mayo Clinic genre looking at the willingness of a woman to take preventive medication at the time of her standardized consult and after receiving results from the the polygenic risk score in addition to a traditional risk estimation model. And this initial pilot was done in non-gene carriers, high-risk women but non-gene carriers. And as you can see, some women's risk estimate went down with the addition of the polygenic risk score, and in others it went up. And as expected, in those where it went down, they were less likely to take preventive medication. And in those that went up, they were more likely to take preventive medication. So it's kind of a no-brainer.
0:29:03.5 HP: We are planning to participate in the second phase of this genre study, the genre 2 study, which will incorporate gene carriers as well, asking a similar question about chemoprevention uptake. So I really think that the future directions in risk assessment are going to be in the field of the polygenic risk score. I think that we'll be doing a lot more with breast density. And the polygenic risk score can also help in prediction of contralateral breast cancer risk. And there were some more recent papers being published on that. It's just... The field is booming and it really can help women make all sorts of different decisions, whether it's screening, whether it's chemoprevention uptake.
0:30:04.3 HP: In some very high-risk women, it may be relevant for surgical decision-making in the preventive setting, but also in newly diagnosed cancer patients. And so how does it fit into risk assessment? SNPs may explain up to 20% of this missing heritability. The PRS increases the discriminatory accuracy of risk models in gene-negative patients, or in those with BUS, or in those that are untested, and may also aid in decision-making for gene-positive patients, especially in those with moderate risk genes. So thank you again for having me here today. And I add a little bit of a clinical bent to the discussion. And I look forward to talking more about it with all of you.
0:31:00.9 TS: Yeah. That was fantastic, Holly. Thank you so much. Let's pause for questions after that great presentation, it covered a lot of ground.
0:31:12.8 ES: I have a couple on the chat, but I'm going to open it up to see if anybody just wants to come off feed and ask before I throw a few out there.
0:31:22.1 Susanna: Edie, I do have a couple of questions. This is Susanna from the Internal Medical Services with Myriad. Thank you so much, Dr. Pederson, for the presentation.
0:31:33.9 HP: You're welcome.
0:31:36.5 Susanna: My question is in regards to how different clinics handle the higher end of the risk with risk score. So our team, we provide medical support to our accounts. And we get a lot of calls from freaked-out patients who get a high number, 60, 70% asking, "What should I do? They're recommending surgery." And the direction that we have been given is to actually say, "This tool is not designed to guide surgical decisions. It's meant to tell if additional screening is necessary." And this seems to contradict a little bit with what you're doing in your clinic. So how do we... Without formal direction from professional societies, we're seeing a lot of variation depending on the clinic. How do we handle that?
0:32:42.1 HP: Yeah, you bring up so many great points with this one question. So number one, we just use clinical judgment when it comes to surgical decision-making, and I would repeat that three or four times, that we don't operate on numbers of any type, and in fact, we don't ever recommend risk-reducing mastectomy in any setting, ever. We only offer a discussion around risk-reducing mastectomy when we feel that the likelihood of the woman getting cancer in her lifetime is, outweighs that she's more likely than not to get breast cancer in her lifetime, that's when we even broach the topic of risk-reducing mastectomy, but I always make it clear to a woman that this is never a recommendation, but is her personal choice, because we do have the alternative of screening and the addition of preventive medication.
0:33:56.1 HP: One thing that I would point out to your providers that, I believe Dr. Slavin may have taught me way back when, is that we are all more likely to succumb to cardiovascular disease than to breast cancer, and so that competing mortality box in the upper right corner of the Tyrer-Cuzick model should really always be checked. And that often brings the numbers down into a range that is less frightening for patients. I think that we don't know exactly where the correct risk level is, and we don't use the numbers to make clinical decisions, and so we use the Tyrer-Cuzick Lifetime Risk Model as it is a family history-based risk model to determine eligibility for MRI screening, and I use the 10-year risk of 5% or greater to encourage a woman to consider preventive medication based on the ASCO recommendations, but we do not use those numbers at all surgically.
0:35:25.9 HP: And that's a huge mistake that's being made, I think, in the general public, and we all need to work together to help educate our providers around that. Check the competing mortality box, use the model for MRI eligibility and discussions around chemo prevention, but we don't ever operate based on risk estimation numbers.
0:35:52.4 Susanna: Okay, that is very helpful, thank you.
0:35:55.0 HP: Sure.
0:35:55.2 Susanna: And I'm excited about what you say about breast density because that can make a huge difference, and as you know, with risk, where we're using version 7, so hopefully in the future, we can be moving in that direction. Thank you.
0:36:15.1 HP: Well, I talked to Alicia and the scientific team about why 7 is used. I assumed it was probably because you didn't have all the information on the breast density, but there is some question around the accuracy of the breast density calculation that's in TC8. You notice how drastically it changes, you know, if you're scattered or heterogeneous and certainly if you're extremely dense, and the modeling was done based on a very small number of patients, really, and so it's an opportunity for us to recreate that work. But I always thought it was unfortunate that we didn't use 8, but now I understand that it actually might be due to a question in accuracy and validation with the addition of the breast density. I do believe it's a huge factor, but I'm not sure that they've gotten it, homed in right yet.
0:37:34.6 TS: Yeah, the only other thing I'll add to this is, yeah, Alicia's working on it, I saw some stuff even from her last week, so I think it's a matter of time before we get it right, but we want to make sure it's accurate and not preemptively rolled out, and then otherwise... I don't know, others might remember differently, but I thought when Tyrer-Cuzick 8 came out, one of the big differences was that it went to age 85 and Tyrer-Cuzick was age 80, but now it seems like the version 7 that's out there goes to 85 as well, so otherwise, they don't seem to have any appreciable differences outside of the breast density, I think there's one other box that's escaping me at the moment, that's...
0:38:19.6 HP: Male breast cancer.
0:38:22.2 TS: Yeah, minor things, so...
0:38:23.6 HP: I don't think they have male on TC7.
0:38:26.4 TS: So the big things are breast density, and then most people have no idea, but actually in Tyrer-Cuzick version 8, if you go to the top in the tools, there's a drop down and you can put a PRS score and...
0:38:36.4 HP: PRS, you can put it in there, yeah.
0:38:39.1 TS: Yeah, so, it is like what we essentially have here is with risk score is really just... The way I kinda think about it sometimes is an enhanced model of Tyrer-Cuzick version 8... Not version 8, Tyrer-Cuzick, I should say. And then one way you can actually play around with the numbers a little bit, we're working hard on trying to get a calculator out so people can recalculate things if family history changes or whatever. A lot of people see referrals in the community and maybe the TRF wasn't filled out correctly or something, and they want to add variables that weren't on the original TRF for the Tyrer-Cuzick risk assessment, but we're working hard to get that out, but in the meantime, you can take the relative risk number that's in the... On the report. Now we put it at the bottom, it's kind of hidden, it's in the paragraph on the second page, sometimes the third page, but it'll be like 1.8 or something like that.
0:39:38.4 TS: That's the relative risk of the polygenic risk score component itself, and you can plug that into Tyrer-Cuzick. And talking about Alicia and Tyrer-Cuzick version 8 and talking about Alicia and Sasha, they don't think it would... They think it should be actually fairly accurate unless there's massive swings for some reason on the reported versus the updated breast cancer family history, because we do do some... We have to control for the double counting, for the family history component that's kind of inherent in polygenic risk scores themselves and then in the Tyrer-Cuzick model, and so there's a little tweaking of that final polygenic relative risk score at the end that has, it's kind of been based on family history.
0:40:20.3 HP: Well, I think it's kind of interesting about the double counting that you refer to, and I think that in some ways, we probably need to look back at atypical hyperplasia and LCIS in that same way, because with the large benign disease database that Mayo maintains and studies, they showed that the addition of atypia to a person with family history, although the models add it together, it really should stay where it was, and so I think you're double counting with atypia... I think atypia and LCIS are probably part of the spectrum of having family history, it's just developing breast cancer, and so, that also is an issue.
0:41:19.9 TS: Yeah, we have that as an exclusion right now, that's why when we... If you look at some of those papers like that 2021 use paper, you'll see kind of a top-off of Tyrer-Cuzick around like 45% when we do our calculations, 'cause we don't have ADH or LCIS, which are really kind of the variables that drive it above 45%, where you'll see risk score kind of have a spread out to 70% or 80% sometimes. It's a good question.
0:41:49.0 ES: So let's talk a little bit about health-reported ancestry and how that is or isn't used. Holly, you gave a very nice kinda overview of how the 56 ancestral SNPs were chosen, but how does that coincide with self-reported ancestry?
0:42:09.6 HP: Well, you know, I think that a lot of our impressions of our ancestry are incorrect, and I think that that's being shown in so many different ways, and people are fascinated by their ancestry and are using probably a lot of the same ancestral SNPs with these ancestry-related tests that we are using in our genomic ancestry determination, but I think that we don't know as much as we thought we knew about our backgrounds.
0:42:56.0 ES: What about, you talked about the three major kind of categories for ancestral SNPs, kind of determination of somebody's just basic ancestry. I'm assuming that hits the majority of people that live in the US, but what about those that may fall outside of those three? Maybe TJ can answer this, are we looking at anything at Myriad outside of those three, or can you just hit on those that maybe not fall into those three groups?
0:43:26.7 HP: There are some plots that Alicia can share. Is Alicia on? I don't think she's on.
0:43:33.2 TS: I don't think so, yeah.
0:43:35.6 HP: She's not on. So there is overlap, and she can explain it best, TJ can probably explain it pretty well as well, but just like when you add more SNPs, it's not necessarily better after a certain point, the incremental difference is really not significantly meaningful. I think that the way that they looked at the ancestral composition, it was similar where there were other... There certainly are a lot of other influencing areas, but if you take those three, you pretty much account for the genomic ancestry, but TJ, would you speak to that a little more?
0:44:31.1 TS: Yeah, yeah, sure. So the way I have been explaining it the last couple of weeks, is thinking about it in my own brain and kinda how to get it across to people is, for each one of the breast cancer SNPs, we're essentially trying to see where in the world it came from, so if you think about it that way. And a good way to do that is to know where in a perfect world, you'd know for each breast cancer SNP, you'd have some SNP in the region that you knew came from this ancestry or that ancestry or that ancestry, and then you could weight it, you'd have weight information across 20 ancestries, for instance, on breast cancer risk based on that particular SNP in the literature. We just don't have that.
0:45:14.9 TS: So essentially, picking the most homogeneous source populations of the world, is a really good way to just calibrate the model, and Alicia has looked at all self-reported ancestries on our test request form, even though we haven't published this, and it fits very well, like the model works very well, whether it's Middle Eastern, Native American, I think Pacific Islander is in there. Everything has really shown a good model fit. So, we don't have any concern that it's not going to work across all ancestries. We'll continue to improve on this. This is just a version in the path towards getting more and more precise risk estimators.
0:46:00.6 TS: If you think about like BRCA1 and 2, or Lynch syndrome, and all these things, over the years, you'd figure out ways to improve. "Oh, these large rearrangements. Okay, I guess we have to look at these. Oh, [0:46:12.0] ____ inversion for NS2-H2." "Oh, I guess we gotta look at this." And that's no difference here, it's a continuous learning process. What I will say is, I think that the core concept of adding background ancestry genetics to determine the breast cancer SNP weighting is really innovative. Like, hats off to Alicia and Sasha and Jerry and team, because that really, I think, will take polygenic risk score analysis to the absolute next level.
0:46:45.9 TS: And so we have a publication under review right now, it's an extremely good journal, [chuckle] we'll see. Fingers crossed. And Holly's part of that as well. And so hopefully, we can get it, all this work out into the literature soon. But it will change, I think the way people even approach doing these kind of ancestral SNP studies from here on out.
0:47:06.4 ES: The way I'd like to look at it is, historically we've used risk models. I used to use them in practice every day as well and I know Holly uses them. It is a tool that we can use to help hone in on what this particular person's risk likely is, and it's not perfect, and it's not exact. And I think PRS and I think risk score, combining those Tyrer-Cuzick historic factors along with the PRS, is a refinement of risk and it helps us [0:47:46.2] ____ a more and maybe accurate and appropriate picture of what that patient's risk probably is. But nothing is exact and nothing is perfect, but it is a refinement of risk above and beyond what Tyrer-Cuzick can offer.
0:48:02.9 ES: And I want to reiterate something that Holly said, that [0:48:06.3] ____ brought up, that patients are calling in saying that they were recommended to have a mastectomy or recommended to have surgery. We don't even recommend a risk-reducing mastectomy in those with... At the highest inherited risks with the BRCA mutation. So, the only time we're recommending surgery is bilateral salpingo-oophorectomy because we don't have good screening tools to identify early ovarian cancer. But we have excellent tools with MRI and mammography to screen and early detect breast cancer. So, I just wanted to reiterate that, that there is no genetic mutation even, that has a recommendation for mastectomy.
0:48:53.5 HP: Yeah, I completely agree. And the risk reducing salpingo-oophorectomy is thus far the only intervention that has been shown to reduce mortality in this group of women. The way I look at risk modeling until now, [chuckle] until we've incorporated actual genomic information, it is not really so much to predict risk as it is to be used for inclusion in clinical trials to reassure patients that are not at high risk, to help convince patients that they might consider preventive medication based on thresholds set forth by societies like ASCO, and to qualify them for MRI screening to enable enhanced surveillance in a woman who you believe to be high risk.
0:50:00.8 HP: I don't use the risk models in a patient who's... Who has no family history and has a radial scar. The sort of across the board risk modeling philosophy is a little bit worrisome to me. I think that you need to use some clinical judgment in terms of who you're targeting for risk modeling and why you're doing the risk modeling. Because until it becomes genomic and the breast density is appropriately integrated and they account for alcohol and the use of preventive medication and other factors, until it's a really accurate tool, I'm not sure that estimating risk is really at the top of my list of uses for it.
0:51:05.7 ES: Well, TJ, we have about five minutes left.
0:51:09.9 TS: Yeah, let me... Yeah, let me... Yeah, I can walk through this really quick. Can people see this?
0:51:20.4 S?: Yup.
0:51:21.1 TS: Okay, yeah. So, I've been working on this a little bit. 'Cause a lot of people have been seeing the polygenic risk component kind of as a black box. And I'm using CHEK2 I157T as an example. So a lot of people on the call know this is a headache variant, to say the least. [chuckle] It's a common CHEK2 variant, frequent in the population. It's about 1 in 200 European individuals. Very hard to figure out how to take care of it. Some labs call it likely pathogenic-pathogenic, reduced penetrants, VUS, all kinds of things. And how do you work up a family like this? So, you have a breast cancer in a mom, CHEK2 I157T positive, and then has many children, some positive, some negative. And is there a way that you can sort this family out?
0:52:19.8 TS: I'm not going to belabor this, but essentially, SNPs, as we've been talking about. This one is actually not quite a SNP because it's actually a little bit below this 1% population frequency. So it's actually called a single nucleotide variant. It would be similar to like Ashkenazi Jewish BRCA1 variant that sit under 1% population frequency as a whole. This is some work that I'm working on with Erin Mundt, and this is about to be submitted. But we've been looking at, in our own cohort, CHEK2 I157T, and clearly, it sits at a very different risk level than what we'd call even missense variants, likely pathogenic and pathogenic and... Which we also don't call the S428F, which some people are familiar with as well, in CHEK2, a pathogenic mutation. So if you actually look at what we call here as truncating or missense, likely pathogenic or pathogenic variants, they tend to have about a two to two-and-a-half fold increase for breast cancer risk, and CHEK2 I157T is way lower.
0:53:30.6 TS: And so how do you kinda pull this into a work-up for a patient? You know, there's a little bit increased risk. You see it here, it's not zero, it's not crossing 1, so there's some slight risk. It looks like a SNP, is the bottom line. And we know risk SNPs are common and they cause a little bit increased risk for breast cancer, and this is really what I wanted to show, because a lot of people have no idea, unbeknownst to many, CHEK2 I157T, and actually, I didn't even know this before I started, is the highest odds ratio SNP in risk scores.
0:54:02.2 TS: So if you actually look at the Mavaddat 2015 paper, which is where a lot of the SNPs came from for our polygenic risk score for breast cancer, CHEK2 I157 was one of the SNPs, and it does perform with the highest odds ratios. There's BRCA2 polymorphic stop-codon, which some people may have come across here and there. Also, somewhat frequent in the population, there's always kind of been a back and forth of whether there's an association with different cancers. It seems like it has a little bit of increased risk for breast cancer. It's the second-highest risk SNP in risk score.
0:54:38.3 TS: There's an FGFR2 SNP, that's the third highest SNP. This ESR1, this is that Amerindian SNP. This is the Latin America. It's a estrogen receptor. It's actually a protective SNP. But you see the odds ratios are just slightly higher, and that's how a risk score is made. You put in the SNP with the highest odds ratio, you look for model fit, and you have the next one, and that's how we kind of topped out around 93%. So the benefits of this type of test is, honestly, it's the most accurate test you could do for someone with a I157T mutation, because you can actually calculate using an entire Cuzick modeling, with the availability of the polygenic risk score on top, you can get really accurate risk scores now. So if this woman's daughter came back positive for I157T, it would bring in all the other clinical and family history variables which are important to figure out, like how important is that I157T to this whole thing, and then all the other background genetic factors.
0:55:37.8 TS: And you can see big swings in families. Some women, if they have more unfavorable polygenic risk scores, could have a much higher risk score whether they have the mutation or not. So that's some of the benefit of doing this type of testing. I just wanted to show that to folks because hopefully, these kind of things make it less of a black box, kinda ground it in. This really all comes from Mendelian genetics. We're just kind of adding Mendelian genetics on top of each other. Where I really anticipate this whole field going is probably like your risk for breast cancer or colon cancer is X, and it's based on all of these other things, whether you have a mutation, other Tyrer-Cuzick variables or whatever it may be, risk SNPs in the background, all these other things, kind of giving more cumulative numbers, and it may bring in some of these genes that we're always looking at, like NTHL1 and all these things that we're like, "Well, it has a really low risk." They function like SNPs essentially, so I think there will be a day where we start bringing in all that data to really inform the risk of the extent we can...
0:56:48.7 HP: And, you know, I think that that initial graph, which we all see and we all start with, with the pie pieces and 70% of women, we have no idea what's going on. I think when the day comes that we truly can estimate risk, that whole chart will flip, and we'll understand 70% and won't understand 30%. And we understand so little, I think, at this point, and that chart just sort of says it all, so we're moving toward... We're moving in the right direction. And thank you to Myriad and to the work that Alicia and Sasha and Jerry do, and it's a great scientific team.
0:57:43.4 TS: Yeah, for sure. Well, great. Well, we're at time. We're actually over now by a minute. I want to thank you, thank you, thank you, Dr. Pederson, so much for coming on. As always, you killed it. You were amazing. Thank you also... [chuckle]
0:57:54.8 HP: Aww, thank you. Thanks so much for having me. Okay.
0:57:57.8 TS: Yeah, no, this was great. And thanks, Edie, for running the chat, and I look forward to seeing everybody in a few weeks.
0:58:02.7 HP: Great. Bye.
0:58:03.0 ES: Bye. Goodbye.
0:58:03.3 TS: Alright. Bye.
0:58:06.6 ES: Thank you.
[music]