NHGRI Issues Polygenic Risk Scores Reporting Standards

NHGRI Issues Polygenic Risk Scores Reporting Standards
Credit: Peter Finch/Getty Images

A 22-item framework has been published identifying the minimal polygenic risk score-related information scientists should include in their studies. Created by NHGRI’s Clinical Genome Resource’s (ClinGen) Complex Disease Working Group and the Polygenic Score Catalog (PGS), an open database of polygenic risk scores, the authors hope this framework will help promote the validity, transparency, and reproducibility of such scores.

Scientists and healthcare providers have started using polygenic risk scores for assessing a person’s inherited risk for diseases such as Type 2 diabetes, coronary heart disease, and breast cancer. These scores are calculated using the data on which variants tend to be found more frequently in the genomes of people with a given disease. By comparing the number and type of variants an individual has against this data, a polygenic score is generated, and that helps estimate an individual’s risk for specific diseases.

Almost 3,000 studies that discuss polygenic risk scores are listed in PubMed. About 400 of these relate to diabetes, 350 are about heart disease, and more than 500 relate to cancer.

However, despite the rise in studies using polygenic risk scores, researchers have observed inconsistencies in how such scores are calculated and reported. These differences, they think, threaten to compromise the adoption of polygenic risk scores in clinical care.

“A real challenge is that the research community has not adopted any universal best practices for reporting polygenic risk scores,” said Erin Ramos, Ph.D., a program director for ClinGen, deputy director of the NHGRI Division of Genomic Medicine and co-author of the paper. “With the field growing as fast as it is, we need standards in place so we can meaningfully evaluate these scores and determine which ones are ready to be used in clinical care.”

This framework builds off another best practice model called the Genetic Risk Prediction Studies (GRIPS) statement, published by an international working group in 2011. GRIPS placed an emphasis on models that included a smaller set of genomic variants and gene scores. However, genetic risk prediction models have evolved rapidly since then, and are based on a much larger set of genomic variants and more complex methodologies. Also, researchers have not fully adopted the GRIPS framework.

“A renewed emphasis on reporting standards by ClinGen and the Polygenic Score Catalog comes at a crucial time for polygenic risk scores,” said Genevieve Wojcik, PhD, MHS, an assistant professor of epidemiology at the Johns Hopkins Bloomberg School of Public Health, Baltimore, and corresponding author of the paper. “It specifies the minimum information that should be described in a research paper for interpreting a polygenic risk score, reproducing results and eventually translating the information into clinical care.”

Some of the new reporting framework items include detailing the study population and the basis for choosing that population.

“If we are to make these scores available to people around the world, the studies need to define who they are studying and why, in the clearest way possible,” said Katrina Goddard, Ph.D., director of Translational and Applied Genomics at the Kaiser Permanente Center for Health Research, Portland, Oregon, who also co-authored the paper. “Without that transparency and reproducibility, efforts to use polygenic risk scores may be undermined.”

The new framework suggests that scientists should explain the statistical methods they used to develop and validate the polygenic risk scores. Without a consistent way of reporting polygenic risk scores, it is nearly impossible to compare the utility of the scores for assessing disease risk in people. According to the new guidelines, researchers should also consider potential limitations of these scores and how clinicians should use the scores in patient care.

“If researchers can follow these guidelines, it will be more straightforward to evaluate published polygenic risk scores and decide which ones are a good fit for the clinical setting,” said Michael Inouye, PHD, director of the Cambridge Baker Systems Genomics Initiative, UK, and co-senior author of the paper. “For diseases such as breast cancer and many others, we will be able to responsibly place patients in different risk categories and provide beneficial screening strategies and treatments. Ideally, in the future we will detect risk early enough to combat the disease effectively.”