Human orthologs were obtained using the One2One R package that is incorporated in the MAGMA_Celltyping R package [8]. SNPs were annotated to the human genome (hg19, version 33) of the GENCODE project [30]. Binary annotations files were created for each cell type, containing 11 sub-annotations. The first sub-annotation contained SNPs that mapped to genes without a human ortholog (1 = SNP belongs to a sub-annotation). The other ten sub-annotations represented the SNPs in specificity deciles for a particular cell type in increasing order. These specificity deciles were obtained by restructuring the specificity metric Sg,c, described in the Methods section “Overview of cell type enrichment analyses” using the “prepare.quantile.groups” function in the MAGMA_Celltyping package [8]. LD scores were then calculated for each annotation file using a 1 centimorgan (cM) window, 1000 Genomes Project Phase 3 files [31], and restricted to 1,217,311 Hapmap3 SNPs. For each summary statistics dataset, we generated munged summary statistics by applying previously described quality control steps [22] (Supplementary Methods), implemented in the LDSC “munge_sumstats.py” script. Finally, SNP-h2 was partitioned, using the munged summary statistics, 1000 Genomes Project
