A framework for exploiting horizontal pleiotropy. If a single nucleotide polymorphism (SNP) being used as an instrumental variable (IV) appears to be an outlier in an MR analysis, then it is hypothesised that it is additionally associated with the outcome through another trait. Searching through genome-wide association study (GWAS) databases to find candidate traits, then modelling the original association between the exposure and outcome accounting for these candidate traits allows a) the detection of new putative exposures and b) a reduction in heterogeneity and/or bias in the original estimate.
This method depends on the ability to accurately estimate the effect estimates on the path through candidate traits and can be automated using available GWAS databases such as MR-Base. However, these databases do not have complete coverage of the human phenome and therefore unlikely to be able to completely account for the heterogeneity.