(iStock)
A Vanderbilt Health-led team has been named a winner of the National Institutes of Health (NIH) Replication Prize for developing a tool that allows researchers to gauge the quality of data in large DNA biobanks by measuring how reliably the data reproduces previously published genetic findings.
Team PGRM was among 15 teams and individuals selected as “Replication Exemplars,” recognized for integrating replication into their standard research practice. NIH announced the winners May 13. Winners will receive a cash prize of up to $50,000.
In its initial release, the team’s tool, the Phenotype-Genotype Reference Map (PGRM), included a curated set of 5,879 phenotype-genotype associations drawn from 523 genome-wide association study (GWAS) publications, standardized to support replication studies across biobanks. It was described in The American Journal of Human Genetics in 2023.
DNA biobanks, which link genetic data with electronic health records for large populations, have become central to genetic discovery, but the field has lacked systematic ways to evaluate whether the medical record data being analyzed accurately reflects the diseases under study. The PGRM addresses that gap by allowing researchers to check how well a biobank cohort replicates a broad set of established genotype-phenotype associations. A low replication rate can flag data quality problems before downstream analyses proceed.
The PGRM was validated across four independent biobanks — BioVU at Vanderbilt Health, the Michigan Genomics Initiative, UK Biobank, and BioBank Japan — and has since been used in published analyses of All of Us, the Taiwan Precision Medicine Initiative and other large cohorts. It is available as an open-source R package on GitHub.
Team PGRM includes Lisa Bastarache, MS, Sarah Delozier, Jing He, MS, Adam Lewis, Robert Carroll, PhD, Jacob Hughey, PhD, and Josh Peterson, MD, MPH, of Vanderbilt Health, along with collaborators from the University of Michigan, the National Human Genome Research Institute at NIH, and Stanford University. The work was supported by the NIH under awards R01LM010685 and UL1TR002243.
