We analyzed two independent sets of primary liver tissues at the University of Chicago (UC; n = 206) and University of Washington (UW; n = 60). We genotyped both sets of samples using Illumina SNP arrays (quad-610 and 550 k for UC and UW, respectively); to improve mapping power [30], [28] and replication ascertainment, additional genotypes were imputed using HAPMAP reference genotype panels (see Methods). Gene expression levels were analyzed using Agilent (UC) and Illumina (UW) expression arrays. We considered the UC liver collection as a ‘discovery’ set and used as replication panels the UW collection and a published set of liver eQTL data (Merck; n = 266) [31]. However, we note that the conclusions drawn below were robust to the choice of a ‘discovery’ set (Figure S1). All samples analyzed across all three studies were unique. Microarray expression probes from both platforms were remapped to RefSeq gene models to aid in cross platform comparisons. A total of 14,703 RefSeq genes were surveyed in the UC reference study while 11,245 RefSeq genes were present on all three platforms. We have made these data and results publicly available through the GEO and SCAN databases (http://www.scandb.org/) [32].
