A recent discussion in our GitHub issue tracker was centred around adding Mendelian gene enrichment as a feature of GWAS locus prioritisation in Open Targets Genetics (see issue #2103).
We thought the ideas shared might be of interest to the Community, so we are relaying the discussion here.
A user suggested we could use a similar approach to MendelVar from Sobczyk, Gaunt, and Paternoster (2021), to prioritise genes near GWAS hits based on enrichment of Mendelian diseases for those genes. The user further suggested that we could substitute the European Variation Archive (EVA) for Disease Ontology and ClinVar using the MendelVar method. (Note that since Mendelian disease mutations in EVA are almost entirely from ClinVar, this would not drastically change the source.)
As part of the work for our recent article on the Open Targets Genetics Locus-to-Gene pipeline, we used the MendelVar method to get Mendelian disease gene annotations near association peaks for a set of GWAS, and our analysis showed that they do tend to have high L2G scores.
Mendelian disease evidence would be valuable to identify causal genes at common disease loci. For Open Targets, the key challenge is to incorporate such information in a systematic way. This would require careful matching of phenotypic characteristics between common and Mendelian diseases, and is something we intend to explore for future releases of Open Targets Genetics.
The user discussed ways in which this matching might be done based on genetics. They noted that this could create a circular dependency with our Locus-to-Gene method, but also commented:
I can almost imagine an EM formulation for the problem or perhaps a two-stage L2G model where the first uses a simpler score like V2G for the initial matching of GWAS genes to rare disease genes.
Thoughts, comments, or queries? Let us know!