We currently assume that each genetic variant is equally likely a priori to affect gene expression or trait. A straightforward addition to our methodology would consider location specific priors for each variant, which would depend for example on the distance to the gene of interest, or the presence of functional elements in this chromosome region [29]. Our computation of the BF also assumes that, under , the effect sizes of the shared variant on both traits are independent. This could be modified if, for example, one compares eQTLs across different tissue types, or the same trait in two different studies. [30] has proposed a framework to deal with correlated effect sizes, and these ideas could potentially be incorporated in our colocalisation test.
