Seven days of extreme GWAS p-values, 2023, episode 4 of 7, complex traits without an obvious causal gene

The first two associations added to the GWAS Catalog were for bone mineral density from a 2008 study from Brent Richards, Nicole Soranzo, Tim Spector and colleagues, added June 10th, 2008.

In the 5672 days since then, the catalog has added another 567,000 associations which cover an additional 10,070 traits (100 new associations and 2 new traits a day for 15 and a half years!)

Today we’re going to look at some of the complex or non-molecular traits (as opposed to protein or metabolite QTLs).

More specifically, I’ve enumerated the 10 most extreme p-values for complex traits where the causal gene is not obvious.

This is perhaps a bit perilous for me because of course what I mean is the causal gene is not obvious to me. So for some of these I may be simply exposing my own ignorance of the underlying mechanistic or biological explanation.

 Still, as I’ve written before, before declaring something a causal gene I like to understand the biological link from the gene to the measured trait. This can be the molecular path from the gene to the phenotype, typically supported by in vitro work, mouse (or canary) knock-outs, or human Mendelian traits. We’ll look at a couple examples of this tomorrow, with the “explained” complex traits.

With that caveat, here are the top 10 “unexplained” complex associations added to the GWAS catalog in 2023:

For each of these associations I have listed the closest (protein coding) gene. In many cases the closest gene may actually be the causal gene, but I don’t think there is enough concrete evidence to definitely establish any of these. Still, closest gene is not a bad place to start.

The winner in this category this year is the hit at NLRP12 for a neutrophil trait from Parsa Akbari, John Danesh, Stephen Burgess, Adam Butterworth, Nicole Soranzo and colleagues, which was introduced with my monthly run-down in October. One can begin to generate hypotheses of why NLRP12 might influence neutrophil morphology but I don’t believe such hypotheses have been tested or validated.

POU5F1B is implicated in a meta-analysis for prostate cancer published last month from  Anqi Wang, David Conti, Chrisopher Haiman et al. POU5F1B, the closest gene to the lead SNP at rs11986220 has frequently been linked to cancer, perhaps acting as an oncogene, a gene which is generally benign but which can promote tumorigenesis under certain circumstances. This makes POU5F1B a strong candidate for causal gene but I don’t think the full molecular path has been worked out.

Another strong candidate causal gene in this table is WNT7B, a member of the Wnt signalling family. At least 6 Wnt-related genes lie close to GWAS hits for Dupuytren’s contracture, consistent with WNT7B being the causal gene here. Still, there is not, to my knowledge, mechanistic evidence specifically linking WNT7B to this specific disease. Further, although there is a rare disease linked to loss of WNT7B, the gene is involved in so many processes it is impossible to definitively link the phenotype of this disease (Matthew-Wood Syndrome) to the signs and symptoms of Dupuytren’s contracture.

What do you make of my “unexplained” list? Any of these seem well-understood at this point? I welcome your feedback.

Tomorrow

Check back tomorrow when I will walk through some examples of complex traits where I think the causal gene has been well established.