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 Phase 3 minor allele frequency files, and both the 1000 Genomes Project phase 3 baseline model and all sub-annotations as independent variables. For the regression weights, we used the LD weights calculated for HapMap3 SNPs, excluding the major histocompatibility complex (MHC) region (chr6: 25–34 Mb) using the “overlap-annot” to account for SNPs grouped into multiple deciles. In addition to the settings described above, we performed sensitivity analyses, including removing the HapMap3 SNPs restriction, using only SNPs that pass a genome-wide significance threshold, changing the software version, and changing the reference genome version to determine differences in cell type enrichment results for SCZ using the KI dataset. To allow comparison of all enrichment methods, cell type enrichment figures show the P value associated with the most specific decile for each cell type as not all methods provide an enrichment score. Methods for MAGMA, DEPICT, and cell type enrichment analyses using additional mouse and human scRNA datasets are outlined in the Supplementary
