Fallacies and Pitfalls in Genome-Wide Association Studies*

Since the first genome-wide association study (GWAS) identifying variants associated with myocardial infarction was published over 20 years ago, GWASs have emerged as a powerful tool for exploring the genetic basis of complex traits. To date, hundreds of thousands of statistically significant associations have been reported across thousands of human phenotypes. Nevertheless, the design, implementation, and analysis of GWASs remain complex, and the results are easily misinterpreted. Common mistakes include 1) assuming that variants with the strongest statistical associations are causal instead of correlative, 2) believing that associated loci act through nearby genes, and 3) overemphasizing the contribution of individual loci to the total variability of particular traits. Clinical assays have been designed using the results of GWAS that rely on the contribution of such erroneous data interpretations to predict clinical phenotypes, reactions to medications or foods, and/or propensity to develop diseases. The failure to recognize these errors due to fallacies in logical reasoning and statistical inference presents problems for both the scientific community when the wrong targets may be prioritized in future research studies, as well as for communication with the general public when our understanding of the genetic basis of important traits may be misrepresented and overstated. Here, we review statistical data quality, analysis, and meta-analysis, of GWAS results with an emphasis on accurate and reliable interpretation. Placed in the appropriate context, GWASs enable genome-wide discovery of loci associated with diverse traits, but they constitute only a first step towards understanding the biological mechanism(s) underlying the observed associations. Scientific elucidation of these biological mechanisms must be required to establish causality with biochemical and pathophysiological explanations for any putative statistical correlations.