In case you missed it, you should go back and read the recent article by Visscher et al., “10 Years of GWAS Discovery: Biology, Function, and Translation” in The American Journal of Human Genetics. It’s an engrossing summary of (recent) scientific history, recapping the evolution of the GWAS, and highlighting ten-years of progress made in the utilization of SNP-based GWAS designs. But in their concluding remarks, the authors pose a daunting question: “Does the GWAS have a future?” And their lengthy defense of GWASs leads me to ask a higher-level question of my own:
“Why does GWAS need defending?"
We at Chrysalis have lived through, and in some cases, have been at the center of companies that made GWASs practicable. For a lot of us in the armchair-genetics community, the crux of the issue is well summed up in the authors’ commentary: “The relationship between sample size and number of risk loci detected [in a GWAS] varies between traits, but all show a sharp increase at a critical sample size. To date, there has been no trait with evidence of a plateau of the number of risk loci discovered with increasing sample size.” Shame on all of us to expect biology to be well behaved! The lack of evidence for a plateau causes the needle-in-a-haystack metaphor to quickly disintegrate. The haystack is, in fact, full of needles! And the more you roll around in the hay, the more you get pricked. Unfortunately, despite a steady increase in sample size, studies on most traits have failed to demonstrate a plateau in the number and diversity of associated pathways as well as in risk-associated genes and variations. I have attempted to capture this notion figuratively, below.
This prediction is both edifying and deeply troubling. On one hand, it suggests that the ever-expanding flood of loci with weak effect sizes has, and will continue to open our eyes to new pathways involved in disease, expanding our understanding of the chemical and structural interconnectedness of human systems. On the flip side, the graph above is a brutal insult to the practice of reductionism. It suggests that the genetic bases of disease risk (albeit not disease causality) cannot be easily compartmentalized, at least not by common variants. And it suggests that using a GWAS to search for a potential drug target in a molecular pathway of great mechanistic effect is a doomed endeavor from the get-go. But for the exceptional cases like those described in the article by Visscher et al., GWASs have largely failed to identify singular pathways of great mechanistic effect. More critically, SNP-based GWASs have failed to help to distinguish those few pathways that are mechanistically important from an overwhelming number of pathways that aren’t.
But let us not waver in our cause! For SNP-based studies are but one of many lenses available in the search for actionable insights into disease mechanism. The historical drivers for the broad adoption of SNP-based GWAS designs were almost entirely economic, and there are new technologies and experimental designs that will give us both economic and mechanistic levers to take us beyond the limits of traditional SNP-based association studies. I want to give a shout out to those those embattled scouts, those biomedical light-brigadiers, deep in the trenches of molecular biology, in the search for a fundamentally better approach. The following figure suggests four opportunities to transcend the GWAS stalemate.
So what do you think? Which technologies and study designs are most likely to yield a more focused, more mechanistic, more actionable view of human disease? Share your feedback, predictions, and insights in the comments section below.