The dataset displayed here (test.parquet) represents the ~10M variants used for S-LDSC in hg38 coordinates (a tiny fraction we couldn't liftover are marked with pos = -1).
We include scores for the 3 GPN-Star models (mutation-rate adjusted minus entropy, higher -> more functional).
