We used the variancePartition package18 in R to calculate the proportion of variance in RNA expression explained by known covariates such as age, gender, RIN and PMI, using the variancePartition package in R. The variancePartition18 package uses linear mixed model based statistical methods to quantify the contribution of multiple sources of variation and identify the covariates that required correction in the final analysis. Supplementary figure 1A shows violin plots depicting drivers of variation in gene expression data without accounting for covariates. The figure shows that sequencing batch is a major driver of variation in a large proportion of genes, while RIN and sex have large effects on only a few genes. We used the voom function in the Limma package (https://www.bioconductor.org/packages/devel/bioc/vignettes/limma/) to account for the effect of sequencing batch, RIN, age, sex and PMI on gene expression. After removing the effects of these covariates, alcohol-related phenotypes explained the largest proportion of the remaining variation in gene expression (Supplementry Figure 1B).
