DNA Today: A Genetics Podcast

10 of 300 episodes indexed
Back to Search - All Episodes

#168 Polygenic Risk Scores with Giordano Bottà

by Kira Dineen
January 21st 2022

Enter a giveaway on our social media! Win free enrollment to a 3-hour course in the Allelica PRS clinical academy covering the research behind PRS to ... More

how is it that we find ourselves surrounded by such complexity. We're all made of. Hello, you're listening or watching DNA Today. We are a genetics podcast and radio show. I'm your host here, Deneen. I'm also a certified genetic counselor practicing in the prenatal space. On this show. We explore genetics impact on our health through conversations with leaders in genetics. These are experts like genetic counselors, researchers, doctors and patient advocates. My guest today is Giordano Botha. He is a biologist and bio informatization. Giordano earned a PhD in public health. He has extensive experience in analysis of large genomic data sets, including a publication in nature. He is a co founder and Ceo of silica which created a software to help clinical genetics labs to perform apologetic risk or analysis in this episode we're gonna be exploring apologetic risk score which is something that I've wanted to explore I'd say for a couple of years. Um the show so thanks so much for coming on and sharing your expertise in this area. Thank you for having me.

Pleasure to be here. So let's start out with just defining what is apologetic risk score because I think it's become a buzzword in genetics but some people may not understand like well what does that exactly mean? Like what is apologetic risk or which we might refer to as prs um throughout the interview but go ahead and define that for us. Yeah. So apologetic risk or is a way to assess the genetic risk of developing a disease. We can also differentiate a pathogenic rescore with apologetic score because apologetic score can be used to predict a trait such as height by pathogenic risk score has been built specifically to predict disease risk and is uh weighted sum of genetic variants. And you know, the opportunity score can be uh built using just few variants up to millions of variants. So there is a very large diversity in how we can we can build the apologetic riscorp and the most important thing to remember.

Is there a way that some, because each variant used in this court has a different weight and then we need to multiply the weight for how many leaders at each variant a person has for example for that effectively. So we call it a so there's a lot of data that goes into this because unlike other testing where maybe you're finding oh there's one change in a gene and this could have this effect. And you're looking at one change this, you're compiling so many changes into one to talk about in effect. Um so it's it's very different from a lot of other genetic testing that people think about and we say, oh there's a mutation, this is like many of those combined. So knowing that how apologetic risk scores empowering the next generation of clinical genomics, I mean this is where we're headed for the future. I think once we heard apologetic scores were you know, becoming a thing? People are like, wow, this is really going to change genomics. How do you see it, changing genomics in the next few years, decade.

Yeah. Yeah. I think that many is because is adding a new layer of information to the what we are using uh till a few years ago in clinical genomics because it is adding the information of using polymorphism. So common barriers that are not errors in our genetic code such as the mutation. So the common variance polymorphisms, they do not cause um a prudent to function in a in a in an alternate way or they do not cause they're not errors. So and for example, polymer films are the genetic variants that can code for different eye color or different morphological traits that we have between different people. And so adding this new layer, it's possible to have a much more precise estimate of the genetic risk of developing diseases also to explain the genetic component of the disease.

And until a few years ago this uh this layer of information. So the use of common variance has been neglected and also for a good reason because we haven't had the data to be able to identify the contribution of those variants and also to validate the contribution of this bias. Because it's a extremely important to start identifying the variants and then validate the predictive power of the score that we have built in prospective studies. And now thanks to the very large genomic data sets that are available since just a few years we were able to uh use the results from A. G. Was. So from association studies and model different forces that act on on the genome to identify this course and then validate their predictive performance in clinical data set.

So what kind of conditions can apologetic risk or be calculated for? We've been talking about what apologetic risk score is, how it works. But how can we actually apply that to human health? Yeah. So potentially we can calculate pathogenic risk or to any trait or disease that we have a genetic data because we we can also build a score using a few individuals. But this won't have any meaningful predictive power in a clinical setting. So I would like to stress that there is a problem in pathogenic risk or that is the generalization. And then people talk about apologetically score can is not predictive enough cannot be used into the clinic but without describing the the use case. And for example in the apologetic score catalog. That is a fantastic resources where scientists can upload the apologetic score. There are more than thousands of different score.

But then just few have the predictive power that can add something to what we already do in medicine. Because apologetic rescore is when we want to add the an innovation in technology and especially In the genetic assessment of the disease. Won't understand how much this new technology add to what is already in place, what is already existing some score. They do not add anything because they can for example, identify an increased risk just of 10%, of a disease that has a prevalence of for example, 5%. So knowing to have the 5.5% risk of diseases is not interesting at all. While for other diseases, other diseases we have a pathogenic risk or that can identify that 34444 the risk of disease. And these start to be relevant into the clinic for some conditions. We see that it's changing risk a little bit but not enough to change what we do clinically or anything to change in someone's life with management.

But for others you're saying I mean a three or four increase like fold risk for that. Maybe there is something to do in those cases. Um but I think as time goes on we're gonna learn more about more and more variants that play a role in that and being able to identify people. So I guess that's what's changing over time is just how many variants that we understand with genomics. Perkinelmer genomics is a global leader in genetic testing, focusing on rare diseases, inherited disorders, newborn screening and hereditary cancer testing services support the full continuum of care from preconception and prenatal to neonatal pediatric and adult testing options include sequencing for target genes, multiple genes, the whole xOM genome and copy number variations using a simple saliva or blood sample. Perkinelmer genomics answers complex genetic questions that can proactively inform patient care and and the diagnostic odyssey for families, learn more at perkinelmer genomics dot com.

Have you done a genetic test and discovered you have a mutation or pathogenic variant as we genetic counselors call them. One of the next steps is to inform your family their risk of also carrying the variant, telling relatives about their risk and save lives. But these can also be tough conversations to have. Luckily connect my variant offers resources to help including ways to communicate results with suggested language. Whether you want to talk to your aunt or email, a third cousin, connect my variant can help you are not alone, connect my variant can connect you with others who have the same exact variant growing a network of people with the exact same variant who know about it. Can help your family members understand the impact of your genetic variant and the steps for them to get tested and to prevent disease. If you want to discover where your variant came from and learn more about what it means for your family, explore and sign up at connect my variant dot org again, that's connect my variant dot org. Are you interested in the rapidly growing field of genetics and want to learn more about clinical genetics, molecular genetics and laboratory science. Then you should check out the genetic assistant online training program at john Hopkins University School of Medicine by taking part in the program.

you'll be joining both national and international learners with the same passion for genetics interact directly with your johns Hopkins instructors and fellow students throughout the program. Applications are closing for the spring cohort but there are still spots available for summer and fall 2022. For more information, use the link in our show notes. If you're listening to this through a podcast app, you can also access the link at D. N. A podcast dot com. Who would you say is the target market for this test? Who can benefit the most from poly genic risk scores at this point? Mhm. Yes, potentially the target market is uh is everyone because we would like to know our genetic risk of developing diseases, especially when in in the world where we live where we have some guidelines that tell us to have a certain screening or to respect some threshold for some lipid levels for example. But this threshold, this uh days of timing of the two initiated screening have been based based on on the average while people have a very nice wide range of diversity and also in how the risk increases over time.

So potentially knowing to have an increased risk of prostate cancer or breast cancer at an early age when we can start to have a uh more frequent screening can be extremely helpful. But then of course human here we need to contextualize because potentially for some score will have some action that can be put in place to reduce the risk for other ones. They we just don't have anything to do. For example for alzheimer risk. So you know we can mitigate the risk but it's not something that There is a clear preventative part to put in place, especially young age while for breast cancer potentially. And uh screening, you know starting at 25, 30 years of age can capture the risk at the very very beginning. So but also there is uh because when we uh we have the new technology in medicine especially we need to be very careful in applying the technology in a place that is not overwhelm the let's say does not overwhelm the practitioners as the physician.

So it's something that can be digested easily. So screening everyone potentially is the right thing to do for certain diseases. But the adoption into the clinic can be uh not uh not be facilitated that with you know, test everyone approach while we can identify some population that will need the more urgently screening by apologetically score this for example applies in a coronary artery disease. Um the prevention. So the heart attack prevention because the cardiologist, they already know that there are some people that are at intermediate risk and they do not know if the best thing to do is to start treatment or even like a more aggressive diet or also like studying intervention or to wait a little bit to see how the risk we progress and that is where apologetic risk or can remake the difference because we can let the physician start the therapy earlier or let them wait a bit more.

That's a fantastic point of for those people in the intermediate range saying, okay well you know are you high risk you not whereas apologetic risk or gives that extra data point or many data points I guess so that physicians, healthcare providers say oh now this does put you into the high risk category. Or now actually we see you're a little bit more on the low risk and to be able to change that management. And as you said, I think breast cancer is a great example of that as well because people may start screening like mammograms, breast M. R. I. S. Etcetera earlier. You know, certainly sometimes I'm sitting with patients and there's a family this year breast cancer, I'm not a cancer genetic counselor but you know certainly comes up with family history and you know, information like this could help patients um In that position to say, okay should I start this early? Should I look at prevention for breast cancer And even Alzheimer's that you mentioned um you know, for that right now there's nothing we can do. But certainly that's probably gonna change in the future as we learn more about you know, more conditions in the neurology space and even people planning their lives if they know they're at a higher risk.

Maybe they'll have different plans maybe they'll you know have their finances be a little bit different if they can see that. So I think even outside of the scope of like direct additional intervention there's also that just global like life's you know scale I guess looking at how this process actually works because we see how important it is and how useful it can be. How does it work? If someone has say they've already done genetic testing they have a genotype how does it work? Can they upload it? What does the process look like for healthcare providers? Because this is a clinical test especially where we work with the clinical genetics lab that they first of all they develop an L. D. T. So it's elaborate developed test. So there are in a clear cup environment so there are a series of validation to perform and depends on the technology used if it's a micro array or a low coverage sequencing. It was also interesting part of prs that we do need uh thorough assessment of the genome because we cannot rely only on exams because a majority of variance are inter genic region in uh regulatory region like in promoter and enhancer.

So we really need to have an understanding of of the genome also non coding regions and then so the kids can be used. The low coverage genome sequencing sequencing the genome A. At one X. Coverage rather than 30 X. This decrease the price is a lot that makes this test affordable for almost everyone. And once the preferred technology has been chosen there is a set of validation process to put in place and then the lab upload the genetic data, the raw genetic data and then our software performed the entire pipeline the up to the reporting. Usually we work with clinical genetics lab to customize the report because each is on a tone of voice. So they want to have the report for the for just for physicians or one report for decision and one report for the individual tested. But the main steps in performing apologetically colonize is is a the course the quality control of the road data and then importation because we can impute common variants.

So since we use common variance we can in for without the variance. Even using panels that can allow the assessment of uh of the genome at like millions of variants using this un expensive technology. Do you want to connect with other people who have the same genetic variant as you? You should check out connect my variant, it's an online resource that allows you to do just that connect my variant also provides different avenues of informing your family of possible inherited risk of disease. This includes helping find where your genetic variant came from and finding distant cousins that may also be at risk. The university of Washington has supported the connect my variant project in an effort to help patients and families understand where their genetic variants come from? Check it out at connect my variant dot org. Are you a student seeking out genetic assistant positions? These vital positions aid with clinical and research patient communication, data entry, genetic testing, coordination and administrative tasks. Therefore training is key.

We recommend the genetic assistant online training program at johns Hopkins University School of Medicine. This online program provides knowledge and skills to learners considering genetic assistant positions or those recently hired to these roles who need job training. The program consists of 2 10 week instructor led courses. All you need is a basic understanding of science, particularly biology. At the successful completion of the program, learners will receive a certificate of completion from the john Hopkins School of Medicine and the Mcusic Nathan's department of genetic medicine applications are closing for the spring cohort but there's still spots available for the summer cohort starting june six and the fall cohort starting september 12th. A limited number of partial tuition waivers are now available. To offset the cost of the program. Don't wait check out the genetic assistant online training program at john Hopkins University School of Medicine. Now click the link in the show notes for this episode. For more information, you can also access the link via the blog post for this episode available at D N. A podcast dot com. Did you know perkinelmer genomics was one of the first laboratories to offer whole genome sequencing on a clinical basis?

Whole genome sequencing can maximize clinical diagnostic yield for patients with a turnaround time of six weeks. Perkin Elmer's whole genome sequencing task is designed to provide access to additional valuable information compared to an X. Um Perkinelmer genomics provides one of the world's most comprehensive programs for detecting clinically significant genomic changes. Perkinelmer genomics delivers knowledge that can empower health. Perkinelmer is a global leader committed to innovating for a healthier world, join their mission at Perkinelmer genomics dot com. And also stay tuned for our interview with Perkinelmer where we're gonna be exploring the power of whole genome sequencing. Also teased that my guest for this episode is a world renowned geneticist. So you're not gonna wanna miss this one. But while you wait for the episode head over to perkinelmer genomics dot com to start discovering all perkinelmer has to offer. So when it comes to sequencing with that you mentioned you know there's different levels of it where sometimes we're sequencing an area of the whole genome multiple times that were really confident on the the sequence of it.

So like the letters of the genome. And you said that that can be expensive the more you do it obviously. Um so if you're doing it less, does that affect the accuracy of the sequence? If you're doing that's called the depth right. When you're doing either one time or 30 times. Does that affect how accurate it can be if in not in this case because we leverage the genomes of uh of the community, let's have the population. So because since we have a very large databases with many genomes sequenced at high coverage, we can reconstruct the part that we are more uncertain because the each each generation, the genome has some variants are associated between one to the other one. Because recombination, it happens just know in one chromosome at the time. So it does not break the entire a prototype.

So the genetic variation are associated among them. So there are statistical way to reconstruct this haplotype with very high confidence. But that's why because we assess common variance, we cannot assess a rare variant because what we do not have the representation in this reference panel, while with the common variance, we can identify exactly that haplotype that's been passed there and is circulating among people. This is also very fascinating because let us understand how related we are to each other in terms of human beings, right? This is how the ancestry genetic testing works as well. You're kind of looking at the same same type of testing that we've been talking about with like um polymorphisms and and those variants. And speaking of that for people that are of non european descent, there's a huge divide in terms of how good we are at giving them genetic information that is equal to people of european descent?

Um So as you said like reference genomes and like the our data sets are primarily of european descent. So when it comes to people that are not there's not as much to compare it to. So you do have this disparity there um that we've highlighted on many episodes of this show. How is silica addressing this in terms of like apologetic risk or because certainly that that has an overlap because you're looking at lots of common variants that we've been talking about. Um You know this is this is a problem in the in the field overall. Um But with a little aka how how are you addressing it? Okay so first of all we validated the scores in each of the measure ancestry. So in african station siltation and latin americans and doing this validation allow us to understand what is the risk, confirm by each percentile of the pathogenic rescore.

And so we can provide the calibrated the apologetic risk also calibrated means that the risk that we uh we provide return to an individual is indeed the the risk corresponding to that ancestry. So we have ancestors specific risk prediction. And so this is the first part validating what has been built with the european data set in non european populations and understand the the risk prediction if it's meaningful or not if it's possible to identify individuals at elevated risk that are that is useful to to use into the clinic or not. And then we are also applying and developing new models that um what they do is to try to understand what are the council barry ins and these are called fine mapping methods. And this method that are extremely important because the main problem is that uh to identify pathogenic risk score, we use G was to train the model and this was, they rely on tagging variants.

So usually the largest signal that we identify is not the council variance but is the variance that is calling linkage with the cows. Also the variant associated with the council because often we do not have genotype, the council balance because also the and the ones that are not based on war genome sequencing, they are based on McGRH array and then importation. There are some limitation on that but also because there is a lot of noise. So since the balance are associated it's difficult to capture the cows. And so what happens that when we uh we apply pathogenic risk or develop with the Europeans in Europeans, we can use the tag invariants because they're extremely associated with the council. But when we try to apply in other ancestry, the linkage disagree, bring pattern. So the association pattern between variants changes.

So the variance that was a linkage with the cows away is not a linkage anymore with the cows away in another population. So we addressed at the noise. This is extremely important to uh to assess to understand if we wanna develop apologetic risk or that are portable between different populations and then what we uh like to tell us that we do not believe that the way to go is to have just one apologetically score that performs in all the different answers. Because we never have something to perform well if you do not uh take into account and embrace diversity because different uh answers if they do have different future and we have to celebrate this diversity rather than try to flatten it down. We are all the same. We we are not, we have some differences. That's the beauty of any a living creature. So we do have to address this and identify what is the best model for that particular ancestry also because they there are some differences also in the underlying biology of the disease and the geology of the diseases.

And for example, now we are seeing some risk models that are already using into the clinic, for example, the poor cohort equation that is a model to assess the risk of heart attack and other cardiovascular diseases. They have had the ethnicity of individuals such for example a african or but it is self reported and why they are increasing the predictive power because they have different betas in the model. So they have different effects. So because each variable each risk factor has a different effect in different ancestral groups. But then this is based on self report the ancestry. And this is a big problem because uh um that's not always correlate with real ancestors that an individual has for different reasons. There's an interesting example of the wisdom trial. There's a clinical trial where they're assessing the use of chemical escorting the breast cancer screening.

We're participants withdrawn the the content content to participate because the she she felt that was not the risk was not addressed in the in the proper way because she she was she at the question which ancestor group she belongs to. She she flagged the the hispanic ancestor group because she had a spanish origin. So she thought that hispanic related with that. But then when a wisdom trial as the new pathogenic risk or for for hispanic and they changed the risk but then they realized that she was from spain and we're not hispanic as intended as a central America. And so she said wait a minute, you change maris 23 times and there's something wrong going on. Just tell that self reported ancestry is really problematic. But what we do and at the first step to build the PRS report is to assess the ancestry of an individual.

And we have two options. One to return the ancestor report, one to do not that for example if someone does not want to know because this can also be considered kind of an incident finding as well because it actually can and they understand many different things on from from the ancestry of of an individual but this is fundamental to calculate the ancestry from a genomic data to build the PRS report. It's interesting because I've even heard of this more happening in other areas of genetics as well of not going by self reported ancestry because as you said, sometimes that is different from their genetic ancestry. Um So I see this as being an area that's really going to be you know, expanding where we're coming up with their genetic ancestry is you know, part of the testing and maybe not reporting that to them because as you said, that could be an incidental finding of maybe they don't want to know that that their family history is not what they thought. Um but that it is medically relevant when we're going to analyze this genetic data before we end I did have one more question that I wanted to learn a little more about.

So there is a percentile of apologetic risk score which I'm hearing is kind of useless and it's more about the absolute risk that matters. Can you kind of explain those two and and how we should be using them before we wrap up here? Yeah. Yeah, sure. Absolutely. Because one big problem is also how to use a quantitative risk factor in in genetics because is a score so people can have like and usually is is normally distributed and we divided in percentile. So the main problem is that the risk conferred by pathogenic rescore th percentile is not correlated with the person tied. So it's not that the the risk increases linearly with the increase of the percentile. So um uh tell in the report that someone has a 82% is that 82%ile of the appearance can cause many problems because first he thinks, okay I have my risk is 80 to uh have the risk greater than 82% of the population.

So it's something that I should be worried about and also 82% can be confused with the actual risk of developing the disease. And but the interesting thing is that the risk increases in many different ways. So for example we can have a jump at the risk at the very end of the distribution. So risk can stay quite flat and then jump just at the 95th percentile. And the absolute risk that can be lifetime risk 10 year or five year risk Increases a crossing the threshold just at 97th%ile for example. So it is extremely important to know the risk confirmed by the prs not just relying on the on the percentile assuming that someone is at the 90%ile. His risk is has there any meaningful clinical application. And also we can think about it like for example if you sort of see the the sides of some fruits, let's say that needs to fit in a in a box, we can have a range of fruits that can fit in a box.

Eventually they are just Little bit differently, 1 to another one. But until they fit in a box, there is not any reason why we need to consider this difference inside relevant. And so it's the same. So since the people they have differential risk up to a certain point where they do not fit, let's say in the box. But they because they have the risk higher than the threshold. So now we need to intervene. But just after the risk across the threshold and that's how how medicine work. We need to put some threshold in the future. There will be like more sophisticated methods that will address the individual differences at each of the risk of distribution. But now we need to be more close. Let's see how the medicine works and how the medicine works with a certain threshold. Putting some threshold. The threshold that we need to rely on are those that are present in the guidelines.

While we develop better model, we we can we need to stick with something that is already used in the practice and I think that's important for health care providers to understand and I know like really focuses on that so that health care providers understand, okay, this is what the data is telling us and this is how you should be using this in terms of clinical management. Be sure to enter our giveaway on social media when a free enrollment to a three hour course in the silica PRS Clinical Academy covering the research behind apologetic risk scores to clinical applications. You can enter by looking for our post on twitter linkedin instagram. It's going to be a video clip from this episode and this has been posted on january 21st the day this episode comes out at nine a.m. Eastern time and the giveaway is going to end on february 4th 2022. So be sure to enter before then. Thank you so much for coming on the show. Dr bogota. I'm just there's so much to polytechnic risk cause I feel like we could go on for like two hours.

Um but I think there's just so much to learn in this space and people can learn more by going to Luca dot com. And you can also learn more about the show by going to D. N. A. Podcast dot com. You can follow us on social media twitter instagram, youtube facebook, connect with us there by searching DNA today and any questions for myself, Doctor bogota. You can send into info at D. N. A podcast dot com and I would really appreciate if you could rate and review on Apple podcast. That's how we're going to reach new listeners and find other genetic nerds like yourself listening. So thank you again for coming on the show. Really appreciate hearing all about apologetic risk scores today. Thank you, Kira. I thank you for listening. Thanks for listening. Everybody join us next time to learn, discover new advances in the world of genetics. We're all made of the same camera we're all made of.

#168 Polygenic Risk Scores with Giordano Bottà
#168 Polygenic Risk Scores with Giordano Bottà
replay_10 forward_10