Despite the publication of 1,455 papers and documentation of 8,062 single nucleotide polymorphisms in the Catalog of Published Genome-Wide Association Studies (GWASs) since 2005, personalized medicine has failed to materialize at the hoped-for rate. A research letter published in the Nov. 14 issue of the Journal of the American Medical Associationmakes the case that large GWASs carry some fundamental limitations despite the hundreds of millions of research dollars spent on them. While effect size is the most important aspect of GWASs, larger sample sizes seem to only elucidate an increasing number of small effect variants. The researchers found that of 1,200 published GWASs only 86 studies (6.8 percent) had odds ratios of greater than 3 at a P value of less than 10−5. Of these studies, half used 300 patients or less. Even very well-established genetic associations have very limited clinical-diagnostic implications. In a meta-analysis of Parkinson disease GWASs, which included 12,386 cases and 21,583 controls, carriers of selected variants had almost 2.5-times increased odds of Parkinson disease, but based on the disease prevalence, this association translates into a lifetime risk for developing the disease of only 0.35 percent, even in the highest-risk group. “These considerations lead to two important […]
Despite the publication of 1,455 papers and documentation of 8,062 single nucleotide polymorphisms in the Catalog of Published Genome-Wide Association Studies (GWASs) since 2005, personalized medicine has failed to materialize at the hoped-for rate. A research letter published in the Nov. 14 issue of the Journal of the American Medical Associationmakes the case that large GWASs carry some fundamental limitations despite the hundreds of millions of research dollars spent on them.
While effect size is the most important aspect of GWASs, larger sample sizes seem to only elucidate an increasing number of small effect variants. The researchers found that of 1,200 published GWASs only 86 studies (6.8 percent) had odds ratios of greater than 3 at a P value of less than 10−5. Of these studies, half used 300 patients or less.
Even very well-established genetic associations have very limited clinical-diagnostic implications. In a meta-analysis of Parkinson disease GWASs, which included 12,386 cases and 21,583 controls, carriers of selected variants had almost 2.5-times increased odds of Parkinson disease, but based on the disease prevalence, this association translates into a lifetime risk for developing the disease of only 0.35 percent, even in the highest-risk group.
“These considerations lead to two important conclusions: risk prediction for an individual usually cannot be derived even from large-scale GWAS data, and sample size is not a quality marker of GWAS per se, especially in terms of clinical relevance,” write the authors, led by Christine Klein, M.D., from the University of Lübeck in Germany.
Complicating current GWAS methodology is the possibility that probands may be more heterogeneous than recognized with phenotypic variants that are difficult to clinically distinguish. Additionally, GWASs of common variants may not delve deep enough into the genetic data.
“Even high-resolution genetic variation will only explain a fraction of the heritability of human diseases and traits,” the authors explain. “Thus, the search is still ongoing for future promise beyond simple genetics with gene-gene and gene-environment interactions, as well as epigenetic effects as important but complex targets.”
