Re-Evaluating Local Genetic Correlation Analysis in the HDL Framework

Local genetic correlation analysis has recently emerged as a way to assess patterns of genetic overlap between phenotypes at a regional level. In a recent report in Nature Genetics, Li et al. present a novel method (HDL-L) that implements such analysis, and use simulations to compare it to an existing method, LAVA. The results suggest statistical shortcomings in LAVA and superior performance for HDL-L. We revisit these conclusions and demonstrate that the simulations in Li et al. exhibit several major flaws that undermine the interpretation of their results. When correcting these issues, we find no support for the conclusions drawn by Li et al., and contrary to their findings show LAVA to have performance generally superior to that of HDL-L.