We used in our analysis gene expression and genotype data from 966 human liver samples. The samples were collected post-mortem or during surgical resection from unrelated European-American subjects from two different non-overlapping studies, which have been described in [16]. The cohorts were both genotyped using Illumina 650Y BeadChip array, and 39,000 expression probes were profiled using Agilent human gene expression arrays. All of the expression data has been normalised as one unit even though they were part of different studies, since high concordance between data generated using the same array platforms has been previously reported. Probe sequences were searched against the human reference genome GRCh37 from 1000 Genomes using BLASTN. Multiple probes mapping to one gene were kept in order to examine possible splicing. The probes were kept and annotated to a specific gene if they were entirely included in genes defined by Ensembl ID or by HGNC symbol using the package biomaRt in R [39]. After mapping and annotating the probes, we were left with 40,548 mapped probes covering 24,927 genes.
