Аннотация:Hundreds of genetic variants have been identified as associated with a spectrum of diseases, but the fine-mapping of causal variants has been complicated by extended linkage disequilibrium (LD) and finite sample sizes.We propose to leverage information between diseases through joint analysis of data from related diseases in a novel Bayesian multinomial stochastic search framework, where prior model probabilities are formulated to favour combinations of models with a degree of sharing of causal variants between diseases.We use simulations and real data examples to illustrate the improved accuracy in comparison to a marginal analysis of each disease.That is, in simulations of two diseases that each have two causal variants, of which one is shared, we find that marginal disease analyses may fail to identify both causal variants for each disease.However, our multinomial framework tends to detect shared variants that are missed by marginal analyses.We jointly fine-map association signals for six diseases and of particular interest is IL2RA, which is known to be associated with several autoimmune diseases, including multiple sclerosis (MS), type 1 diabetes (T1D), autoimmune thyroid disease (AITD) and coeliac disease.Our proposed approach is computationally efficient and adds only five minutes overhead to the fine mapping of individual diseases.
