Identification of SNPs in high apparent LD and statistical retesting with these newly defined clusters did not qualitatively alter the detected cis-eQTL. The benefit of completing this analysis was in reducing undue emphasis on highly correlated genotypes in the data set, thereby more closely following the assumption of independent testing and eliminating weak associations based on a single SNP. Significant expression traits were more highly concordant between different window sizes. In addition, the results were less biologically misleading compared with single SNP-based data. It is likely that, for most significant eQTL, the exact causative polymorphic site was not measured by the genotyping platform and its effect was detected only through a nearby marker in high LD.
