For cell type enrichment analyses using LDSC, we initially adopted the same parameters that were previously described [8]. Additionally, to optimize the cell type enrichment pipeline we tested various settings (Fig. S5, Table S5). The parameters described in the Methods section “Cell type enrichment using LDSC” provided cell type enrichment results that were most consistent with DEPICT and MAGMA. For MAGMA, we found that restricting to HapMap3 SNPs and excluding the MHC region increased statistical power to identify associated cell types, whilst not inflating cell types that were not associated (Fig. S6, Table S6). In addition to MAGMA version 1.08, we also performed cell type enrichment analyses using MAGMA version 1.07b (Table S7). We found that, although cell type associations follow similar patterns using both versions, the updated SNP-wise mean gene analysis model modestly exerts effects on cell type enrichment results, resulting in differently associated cell types (Fig. S7). However, no consistent unidirectional differences in cell type enrichment results were observed. Finally, for DEPICT, we found that not restricting to HapMap3 SNPs increased statistical power to identify associated cell types, without an upwards bias for non-associated cell types (Fig. S8, Table S8).
