The Sequence 3/25-3/31
The In and Outs of Polygenic Risk Scores: Why This Genetic Test is Ready for Primetime but not for Embryos
The In and Outs of Polygenic Risk Scores: Why This Genetic Test is Ready for Primetime but not for Embryos
Polygenic risk scores (PRSs) aggregate the effects of many genetic risk variants and can be used to predict an individual’s genetic predisposition to a disease or phenotype. Using PRSs in the clinical setting to know if a person is at higher risk for a condition caused from the combination of multiple genetic variants, like heart disease or cancer, is long overdue. The problem with integrating PRSs into clinic has been lack of adequate multi-ancestry data and validation. Although various forms of PRSs are available to consumers through commercial platforms, clinical use of Eurocentric PRSs in diverse patient samples risks exacerbating existing health disparities, and there is little long-term data to be found on the outcomes of PRSs.
This week, I’ll discuss the recommendations for clinical implementation of PRSs; both where they are recommended and where they are not.
What is a polygenic risk score?
A polygenic risk score is a number, or a ‘score’ that estimates an individual’s risk for a certain condition. They are used in conditions that are caused by changes in many genes, often coupled with environmental factors. To calculate a score for a particular disease, researchers develop an algorithm that is trained on thousands of genomes using genome-wide association studies (GWAS) data from people who either do or do not have the disease. Variants that exceed a certain threshold of frequency from the GWAS data will be considered for inclusion in the PRS. From there, the validity of the PRS is assessed in a process called ‘discrimination’ in which researchers assess the ability of an algorithm to separate cases and controls. In other words, they look at the probability that a randomly selected 'case’, or affected sample, has a higher predicted risk by the PRS than a randomly selected control. If the probability is high, then the test has a high sensitivity and specificity; it is likely to predict high PRSs for cases and low PRSs for controls. A person’s own score is calculated by feeding their genetic data into the PRS algorithm.
In what setting are polygenic risk scores recommended?
In a recent article, the eMERGE (Electronic Medical Records and Genomics) multicenter consortium published validated, clinic-ready PRSs for asthma, breast cancer, atrial fibrillation, chronic kidney disease, coronary heart disease, hypercholesterolemia, obesity, prostate cancer, and both type 1 and type 2 diabetes.
They began by initially studying PRSs for 23 conditions based on their prevalence in the general population, heritability, medical actionability, their potential clinical utility, and previous performances of PRSs for each given condition. By medical actionability, I mean the ability to take actionable measures to prevent disease. Think: taking a statin to prevent heart disease, undergoing a screening test for cancer.
The research team then evaluated the performance of each PRS across four major ancestry groups: African, Asian, European, and Hispanic. This part of eMERGE’s analysis is key. PRSs tend to be less reliable for minority populations for whom we have less accurate data, and so the predominance of European ancestry populations in GWAS has meant that PRSs resulting from these studies generally work best in those populations. The team was able to acquire such a diverse study population in the first place as nearly half was drawn from the All of Us Research Program, consisting of racial and ethnic minority groups. Another difference in the way eMERGE conducted their analysis: they defined a participant's ancestry based on genotype, rather than by self-reported group identity. After reducing the conditions selected to just 10 that met all the selection criteria, the researchers re-calibrated the PRSs for the diverse set of study data used from the All of Us program.
The researchers validated and verified the performance of the re-calibrated PRSs and then developed a framework for returning PRS clinical reports to providers. The program plans to observe what tests are ordered based on their data, and assess how the PRSs did in the future. The PRSs by eMERGE are ready for clinic and follow-up.
In what setting are polygenic risk scores not recommended?
In a statement recently published by the American College of Medical Genetics and Genomics (ACMG), the group explained why evidence for the clinical utility of preimplantation genetic testing for polygenic disorders (PGT-P) is insufficient.
The first hurdle is getting an accurate picture of an embryo's genes. During IVF, doctors can only take a small sample of cells from an embryo for testing. This limited amount of genetic material makes it hard to get a complete and reliable picture of the embryo's DNA. Even when supplemented with information from the parents' genes, there's still not enough research to be confident that this approach accurately predicts an embryo's future risk of disease.
Another problem is that PRSs derived from GWAS currently perform best for adult-onset disorders. These diseases are often influenced by lifestyle and environment over a person's whole life. As an example, the group singled out coronary artery disease for having one of the most predictive clinical PRS. Using this example, even if an embryo has a low genetic risk score for a condition like heart disease, it could develop later due to poor diet or lack of exercise. This challenges the clinical utility of using PGT-P to choose embryos based on these scores. The PRSs may not accurately reflect the person's chances of getting the disease if it is a condition for which disease occurrence is more often caused by environmental and lifestyle factors than heritable risk factors.
Finally, another significant limitation of the clinical utility of PGT-P is that it's difficult to prove if it actually makes a difference. It could take many years to see if someone develops a disease the test deemed as ‘increased risk’, and there's no way to know for sure if they would have been healthier had they not undergone PGT-P. These kinds of clinical trials will also be very difficult to conduct given the many lifestyle and environmental factors that can influence the outcome.
What’s the takeaway?
The problem with integrating PRSs into clinic has been lack of adequate multi-ancestry data and validation. eMERGE’s approach to calculating PRS using a multi-ancestry dataset is helping to overcome that barrier. PRSs derived from GWAS are clinically appropriate for adult-onset disorders as long as people understand PRSs don’t indicate when or even whether a condition will occur. Importantly, the scores reported by eMERGE still generally perform better on those with European ancestry.
There are still several problems with integrating PRSs into clinic, including the lack of sufficient data on the clinical utility of PGT-P as well as the questions concerning the clinical validity of PRS for embryo selection. Although the evidence base for PGT-P is minimal, research is propelling the science forward, likely initially with prospective trials to demonstrate the near-term safety of PGT-P and eventually with long-term clinical trials to understand PGT-P’s utility.
Newsletter Sources:
Image: https://www.canva.com/photos/MADmulZ7coM/
https://www.nature.com/articles/s41591-024-02796-z
https://www.genome.gov/Health/Genomics-and-Medicine/Polygenic-risk-scores
https://www.nature.com/articles/d41586-024-00502-0
https://theweeklysequence.substack.com/p/the-sequence-219-225
https://www.genomeweb.com/genetic-research/results-emerge-study-edge-polygenic-risk-scores-towards-clinic?utm_source=Sailthru&utm_medium=email&utm_campaign=GW%20Premium%20Weekly%20Monday%202024-03-11&utm_term=GW%20Premium%20Newsletter
https://pubmed.ncbi.nlm.nih.gov/32245787/
https://www.gimjournal.org/article/S1098-3600(23)01068-7/fulltext
https://www.genomeweb.com/genetic-research/acmg-asserts-evidence-insufficient-clinical-utility-polygenic-risk-scores-embryo
