To maximize comparability with the CMC data, we designed an analysis pipeline analogous to that which we used for the CMC RNA-seq processing. We remapped probes to genomic locations of genes using the sequence of the probe (using the same reference genome and Ensembl transcriptome as for the CMC RNA-seq data). For transcripts with more than one probe, we chose the probe with the maximum intensity for each sample (this choice had only minimal impact on results). We retained samples with genotype data so that we could include ancestry as a covariate.We selected covariates based on variance explained in data. The following covariates were used in the differential expression (and eQTL) analysis: Dx, Age of death, sex, PMI, pH, RIN, clustered processing batch, and ancestry markers. We performed differential expression analysis with adjustment for covariates, using linear regression models in limma and identified 2,288 differentially expressed genes at FDR 5%, among which 1,166 and 1,122 were up-regulated and down-regulated in SCZ, respectively.
