Julianna LeMieux Ph.D. Senior Science Writer GEN
GWAS for Common Disease Variants Gains Prominence
A GWAS study focused on polygenic risk scores (PRS) received the type of attention last week reserved for groundbreaking science, including stories in Forbes and The New York Times.
The researchers, led by Sekar Kathiresan, M.D., director of the Center for Genomic Medicine at Massachusetts General Hospital, demonstrated the degree of risk to individuals at the extreme tails of the distribution for five common diseases: coronary artery disease, atrial fibrillation, type 2 diabetes (T2D), inflammatory bowel disease (IBD), and breast cancer. The study, “Genome-wide polygenic scores for common diseases identify individuals with risk equivalent to monogenic mutations,” used well-established techniques for polygenic risk prediction in the largest relevant cohorts available to date (the UK Biobank [UKBB] population).
The goal of PRS is to stratify patients into risk categories based on their genetic mutations. It is unusual for a disease to be caused by mutations in a single gene; these disorders, typically inherited in either recessive or dominant fashion, are called Mendelian. Much more commonly, diseases (including cardiovascular disease, T2D, and brain disorders) are mediated by a collection of common and low-frequency genetic variants, most of which remain unknown. Each variant has a small effect, but, taken together, they could indicate a person’s overall risk. Not only are improved risk assessment tools generally a hot topic, but, the promise of PRS to revolutionize medicine makes it exciting on multiple levels.
Mark McCarthy, M.D., the Robert Turner Professor of Diabetic Medicine at the University of Oxford, explains the significance of using a PRS like this, “Whilst it is true that the overall predictive power for any given disease is modest—for many people with ‘average risk,’ quantifying the exact degree of genetic risk is not likely to be particularly useful—recent thinking has switched to the value of the PRS approach to pick out individuals who have particularly high or low genetic risk.” He adds that, “as this paper shows, many of these individuals have lifetime risk of disease arising from the joint action of many variants of individually small effects, that approaches that of individuals from monogenic forms of disease. It demonstrates that the former group of high-risk individuals are likely to be much more numerous than the latter.”
Dr. McCarthy lends the perspective that “In the first few years after GWAS methods emerged, the small effect size of individual loci, and the limited amount of variance in risk attributable to the identified variants, led to the sense that this kind of prediction would have limited value.” He adds that, “However, as sample sizes have increased, and the amount of individual variation in disease risk that can be quantified through GWAS has risen, the power of this approach has become clearer.”
However, some experts are not convinced that the paper's media coverage matched the importance of the results. Ali Torkamani, Ph.D., director of genomics and genome informatics at the Scripps Research Translational Institute, told GEN that the study published last week in Nature Genetics is “not particularly” a big leap forward in the field of polygenic risk prediction. He adds that the aforementioned paper was “not a methodological advance or even an unexpected result.” Dr. McCarthy agrees that the data were not surprising and that his own group has generated similar data for T2D in their analysis of the UKBB data.
Jose Florez, M.D., Ph.D., chief of the diabetes unit at the Massachusetts General Hospital, who was not involved in the study, tells GEN that the power of this study is in the numbers. “The more people, the more variants, the more separation you can see.”
Dr. Torkamani explains that “although the methods and much of the findings here are previously known or anticipated, they are the first to demonstrate PRS can convey very high-risk at the tails of the risk distribution,” and that the paper is “an important demonstration that clearly actionable disease risk information can be conveyed by polygenic risk prediction.”
However, the usefulness of PRS is still in question and will certainly vary between diseases. Disease risk can be broken down, generally, into three factors: genetic susceptibility, environmental exposures, and lifestyle factors. And, for some diseases, the genetic component is not the most important factor.
Dr. Florez explains that, in the field of cardiac disease, PRS may be a better risk assessment tool than cholesterol measurement. If the PRS can be obtained early enough, a person could avoid the damage caused by exposure to high LDL cholesterol long before it manifests itself in the blood.
That said, he does not see the same usefulness of PRS in his own field of T2D. Dr. Florez tells GEN that he can predict a person’s risk of having T2D with two simple steps: weighing them and taking their blood sugar. In fact, he says that physicians can do as good, if not better, of a job than using a PRS. Moreover, in a disease like T2D the genetics may add a complication that clouds the clinical information. For example, an obese person with a low PRS needs T2D prevention more than a thin person with a high PRS.
Another caveat according to Dr. Torkamani is that there are multiple flavors of PRS. The recent report by Dr. Kathiresan is a “genome-wide polygenic score.” A genome-wide polygenic score “uses a very large number of variants, and makes certain assumptions, which allows the authors to show greater stratification of risk at the tail ends of the distribution. But, the results are less likely to generalize to other populations.”
Dr. McCarthy agrees that the PRS highlighted in the paper is “not likely to optimally capture the phenotypic and clinical heterogeneity within these conditions—that may be better served by developing a series of partial PRS that captures the etiological heterogeneity that contributes to each of these conditions.”
A true PRS, using a sum of risks across loci known to be associated with the disease, is already available today to patients who participate in a research study run by the Scripps Research Translational Institute. Their app, MyGeneRank, has currently 948 users and provides 57 genetic risk markers.
However exciting, the realization of using PRS is still in its infancy and many barriers lay ahead. Dr. Torkamani cites one of them in a recent review as “physician and public education regarding the interpretation of polygenic risk, especially in the understanding of various and dynamic risk metrics.”