Estimating genetic correlation jointly using individual-level and summary-level GWAS data.

With increasing access to individual-level data from genome-wide association studies, it is now common for researchers to have individual-level data of some traits, whereas for other traits only publicly released summary statistics are available due to privacy and safety concerns. Current genetic-correlation methods require both traits to be in the same format. When one trait has individual-level data and the other has only summary statistics, researchers often convert individual-level data to summary statistics and then apply summary-based methods, which is inefficient and can lose information. Here we introduce GENJI, a method for estimating within-population or transethnic genetic correlation based on individual-level data for one trait and summary-level data for the other. Across extensive simulations and real data analyses, GENJI produces more reliable and efficient estimation than summary data-based methods. We further show that more accurate genetic correlation estimation can improve cross-population polygenic risk prediction.