An integrative multi-omics analysis of 16 autoimmune diseases and cancer outcomes highlights immune-cell regulatory mechanisms and shared genetic architecture

BackgroundDeveloping functional understanding into the causal molecular drivers of immunological disease is a critical challenge in genomic medicine. Here we systematically apply Mendelian randomization (MR), genetic colocalization, immune cell-type enrichment and phenome-wide association methods to investigate the effect of genetically predicted gene expression on 12 autoimmune and 4 cancer outcomes. ResultsUsing whole blood derived estimates for regulatory variants from the eQTLGen consortium (n=31,684) we constructed genetic risk scores (r2<0.1) for 10,104 genes. Applying the inverse-variance weighted Mendelian randomization method transcriptome-wide whilst accounting for linkage disequilibrium structure identified 773 unique genes with evidence of a genetically predicted effect on at least one disease outcome (P<4.81 x 10-5). We next undertook genetic colocalization to investigate whether these effects may be confined to specific cell-types using gene expression data derived from 18 types of immune cells. This highlighted many cell-type dependent effects, such as PRKCQ expression and asthma risk (posterior probability of association (PPA)=0.998), which was T-cell specific, as well as TPM3 expression and prostate cancer risk (PPA=0.821), which was restricted to monocytes. Phenome-wide analyses on 320 complex traits allowed us to explore the shared genetic architecture and prioritize key drivers of disease risk, such as CASP10 which provided evidence of an effect on 7 cancer-related outcomes. Similarly, these evaluations of pervasive pleiotropy may be valuable for evaluations of therapeutic targets to help identify potential adverse effects. ConclusionsOur atlas of results can be used to characterize known and novel loci in autoimmune disease and cancer susceptibility, both in terms of developing insight into cell-type dependent effects as well as dissecting shared genetic architecture and disease pathways. As exemplar, we have highlighted several key findings in this study, although similar evaluations can be conducted interactively at http://mrcieu.mrsoftware.org/immuno_MR/.