Most previous studies have used a linear regression (LR) model to associate SNP genotypes with gene expression. We employed an alternative non-parametric method to evaluate the sensitivity of our results to the statistical methodology used. We used Spearman rank correlation (SRC) to perform the same cis- analysis described above. We detected 293, 274, 326 and 363 cis- associations for CEU, CHB, JPT and YRI, corresponding to 783 distinct genes and FDR of 4-5%. Of those, 283 genes were detected in at least 2 populations and 57 in all four. The overlap of SRC with LR was between 77% and 86% of the genes, depending on the population (Table 1). We conclude that SRC performs at an equivalent level as LR.
