As input, mash requires results from a condition-by-condition analysis (effect estimates and standard errors). We used the effect estimates and standard errors for candidate local (“cis”) eQTLs for each gene (GTEx data release 6). These were obtained by performing (univariate) single-SNP eQTL analyses in each tissue using Matrix eQTL19. Expression levels were corrected for population structure (using genotype principal components20) and other confounding factors (both measured factors such as age and sex, and unmeasured factors estimated using factor analysis21), then rank-transformed to the corresponding quantiles of a standard normal distribution. Because these estimates were obtained by single-SNP analysis, the estimated effects for each SNP reflect both the effects of the SNP itself and SNPs in LD with it. Thus, our analyses do not distinguish causal eQTLs from those that are in LD with the causal eQTLs; see Discussion and Supplementary Note.
