Details of MGS subject recruitment and sample characteristics are provided in the online Full Methods (section A1). DNA samples were genotyped using the Affymetrix 6.0 array at the Broad Institute. Samples (5.3%) were excluded for high missing data rates, outlier proportions of heterozygous genotypes, incorrect sex or genotypic relatedness to other subjects. SNPs (7% for African American, 25% for European-ancestry and 27% for combined analyses) were excluded for minor allele frequencies less than 1%, high missing data rates, Hardy-Weinberg deviation (controls), or excessive Mendelian errors (trios), discordant genotypes (duplicate samples) or large allele frequency differences among DNA plates. Principal component scores reflecting continental and within-Europe ancestries of each subject were computed and outliers excluded. Genomic control λ values for autosomes after QC were 1.042 for African American and 1.087 for the larger European-ancestry and combined analyses.
For MGS, association of single SNPs to schizophrenia was tested by logistic regression (trend test) using PLINK10 (link), separately for European-ancestry, African American and combined datasets, correcting for principal component scores that reflected geographical gradients or that differed between cases and controls, and for sex for chromosome X and pseudoautosomal SNPs. Genotypic data were imputed for 192 regions surrounding the best findings, and for additional regions selected for meta-analysis.11 (link) Detailed results are available inSupplementary Datafiles 1 and 2 , and complete results from dbGAP (www.ncbi.nlm.nih.gov/sites/entrez?db=gap ).
Meta-analysis of the MGS, ISC and SGENE datasets was carried out by combining P-values for all SNPs (in the selected regions) for which genotyped or imputed data were available for all datasets, with weights computed from case-control sample sizes. See the companion papers for details of the ISC and SGENE analyses.
For MGS, association of single SNPs to schizophrenia was tested by logistic regression (trend test) using PLINK10 (link), separately for European-ancestry, African American and combined datasets, correcting for principal component scores that reflected geographical gradients or that differed between cases and controls, and for sex for chromosome X and pseudoautosomal SNPs. Genotypic data were imputed for 192 regions surrounding the best findings, and for additional regions selected for meta-analysis.11 (link) Detailed results are available in
Meta-analysis of the MGS, ISC and SGENE datasets was carried out by combining P-values for all SNPs (in the selected regions) for which genotyped or imputed data were available for all datasets, with weights computed from case-control sample sizes. See the companion papers for details of the ISC and SGENE analyses.
