We used S-PrediXcan to integrate eQTL (expression quantitative trait loci) information with our GWAS summary statistics to identify genes of which genetically predicted expression levels are associated with cannabis use26. Briefly, S-PrediXcan estimates gene expression weights by training a linear prediction model in samples with both gene expression and SNP genotype data. The weights are then used to predict gene expression from GWAS summary statistics, while incorporating the variance and co-variance of SNPs from an LD reference panel. We used expression weights for 48 tissues from the GTEx Project (V7) and the DGN whole blood cohort generated by Gamazon et al.56, and LD information from the 1000 Genomes Project Phase 357. These data were processed with beta values and standard errors from the lifetime cannabis use GWAS meta-analysis to estimate the expression-GWAS association statistic. We used a transcriptome-wide significance threshold of p<1.92e-07, which is the Bonferroni corrected threshold when adjusting for all tissues and genes (i.e. N=259,825 gene-based tests in the GTEx and DGN reference sets).
