We subsequently sought experimental support for the functional nature of the most strongly associated SNPs. Therefore, for 14 genes, we cloned (and sequence-verified) common haplotypes existing in the UW liver samples into a customized luciferase reporter vector, and tested the function of each haplotype using high-throughput, transient transfection reporter assays (Table S3; 9 of 14 underlying cis-eQTLs replicated in the UC or Merck samples). For each haplotype, multiple independent vector (mode of 3) preparations were made, and for each plasmid preparation 4 transfection replicates were performed (mode of 12 measurements per haplotype). We analyzed the resulting data using a random-effects model that accounted for both variation in transfection replicates and variation in vector preparations. Our results underscore the need to perform multiple independent DNA preparations to reliably infer sequence-specific functional effects with this system (Figure 5 and data not shown).
