All RNA-seq samples were integrated into a single analysis across the 4 brain banks in order to perform a joint quality control. Retained samples had acceptable values for RIN (mean 7.6, sd ±0.9), intergenic rate (mean 5.7%, sd ±1.9%), total read pairs (mean 4.6e + 7, sd ±1.1e + 7), intronic rate (mean 36.7%, sd ±11.2%), mapped read pairs (mean 4.42e + 7, sd ±9.8e + 6) and ribosomal RNA rate (mean 0.03%, sd ±0.01%) (Fig. 1 and Supplementary Fig. 2). Joint principal components analysis (PCA) of genes on autosomes identified outliers (Fig. 2a). As expected, PCA identified two distinct clusters separated along PC2 that correspond to technical differences in the RNA library preparation. Among other technical differences, samples from HBCC underwent a negative-strand library preparation, while the library preparation of samples from the remaining brain banks was not stranded. Joint downstream analysis must therefore include an RNA library preparation indicator variable as a covariate in order to account for this technical source of variation.
