We estimated European LD Scores from 378 phased European individuals (excluding one individual from a pair of cousins) from the 1000 Genomes Project reference panel using the –ld-mean-rsq option implemented in the GCTA39 software package (with flags --ld-mean-rsq –ld-rsq-cutoff 0 –maf 0.00001; we implemented a 1centiMorgan (cM) window using the –ld-wind flag and modified .bim files with physical coordinates replaced with genetic coordinates as described in the next paragraph – note that a 1cM window be achieved more conveniently using the flags –l2 and –ld-wind-cm in the LDSC software package by the authors). The primary rationale for using a sequenced reference panel containing several hundred individuals for LD Score estimation rather than a genotyped GWAS control panel with several thousand individuals was that even after imputing off-chip genotypes, the variants available from a genotyping array only account for a subset of all variants. Using only a subset of all variants for estimating LD Score produces estimates that are biased downwards.
