Exploiting Gene-Expression Deconvolution to Probe the Genetics of the Immune System
Fig 2
Flow chart of the VoCAL algorithm.
(A). An overview of the analysis. VoCAL takes as input gene expression profiles in complex tissues from a population of genotyped individuals and a reference data containing expression profiles of isolated immune cell types (top box). VoCAL utilizes this input to identify significant 'immune trait associations'—the associations between immune cell type abundance levels and their underlying iQTLs. (B) The VoCAL pipeline. VoCAL proceeds in five steps: step 1—choosing the initial k sets of marker genes; step 2—predicting immune traits (namely, cell type quantities across individuals) based on the reference data and gene expression of the marker genes in a tissue; step 3—mapping the association between each immune trait and each genomic locus. VoCAL generates replicates of this association map by applying steps 2–3 repeatedly using each of the k (non-overlapping) marker sets from step 1; step 4—consolidating the collection of association maps into combined association P-values between each immune trait and each locus. Significantly associated loci are referred to as iQTLs; step 5—filtration of the k marker sets to exclude potentially confounding eQTL targets. If at least one marker is filtered out, VoCAL returns to step 2.