Probabilistic Approach Improves PGS Accuracy by Accounting for Genotyping Errors

Errors in genotyping can contribute to uncertainties in polygenic score estimates but a new study presents a probabilistic approach to account to improve the accuracy of PGSs. In a paper appearing in the American Journal of Human Genetics, a University of California, Los Angeles-led team analyzed data from more than 800 individuals from the Dana-Farber PROFILE cohort for whom there was both low-coverage whole-genome sequencing data and array genotyping data. They confirmed that individuals with lower sequencing coverage had higher PGS uncertainty or errors. But by adding their probabilistic approach into PGS estimates, the researchers improved classification accuracy by 6 percent, as compared to standard approaches. Additionally, using simulated and real low-coverage whole-genome sequencing data, the researchers found that ignoring genotyping errors affected confidence intervals size calibrations of PGS estimates. "Taken together, our results showcase that genotyping errors need to be accounted for in PGS applications for datasets where [low-coverage whole-genome sequencing] is the primary approach for obtaining genotypes," the researchers write.