Using this model, we evaluated the association between the germline genotypes of 816,362 SNP loci and genes within 1 Mb on either side of each SNP locus. In order to further control the false positive rate, we excluded 396 genes with low expression levels (less 10% present call), 85 genes with probes having a known sequence polymorphism, as well as the highly polymorphic HLA region. A Q-Q plot of the raw p-values corresponding to the cis-associations tested before and after adjusting for somatic factors suggests systematic bias in the unadjusted data (λ = 1.29), which is eliminated after the adjustment (λ = 0.94∼1.01, Figure S1).
