NEW YORK (GenomeWeb) – An international team led by researchers at 23andMe have demonstrated the utility and feasibility of using phenome-wide association studies (PheWAS) in population-based cohorts to evaluate drug candidates or predict adverse drug effects.
The researchers used a PheWAS to assess candidate drug targets in the genome, falling near 25 SNPs found through past genome-wide association of common diseases. From phenotypic associations found with data for almost 698,000 participants in 23andMe, UK Biobank, Finrisk, and Children's Hospital of Philadelphia (CHOP) cohorts, and validation testing in more than 337,000 UK Biobank participants, they narrowed in on nine significant associations, including those offering clues to potential adverse drug events.
"Taken together, our study highlights PheWAS as a highly promising, yet largely untapped opportunity to use disease-agnostic cohorts with extensive health information for drug target validation," senior and co-corresponding author Heiko Runz and his colleagues wrote in their study in Nature Communications today. Runz, who was affiliated with Merck when the study was performed, is currently the medical director of genetics at Biogen.
PheWAS have been proposed as a strategy for analyzing drug candidates, assessing target-disease ties, repurposing existing therapies, and finding clues to potential adverse drug effects, the team explained, though more research is needed to explore that possibility in broad population cohorts.
"Despite these benefits, the ability for PheWAS to substantially add to the decision making in drug development is thwarted by the difficulty to obtain and systematize comprehensive genotypes and phenotypes across very large numbers of individuals," the authors explained.
For the new study, Runz and his colleagues sifted through the available literature to find 19 proposed drug target genes. With the help of self-reported phenotypes and/or health measurements for as many as 697,815 individuals from the four population cohorts, including 671,151 23andMe participants, they searched for pleiotropic effects for 25 SNPs near these target genes, previously linked to conditions such as Crohn's disease, body mass index, psoriasis, or type 2 diabetes.
By looking at 1,683 binary endpoints in participants enrolled through the population cohorts included in the study, along with a meta-analysis looking at 145 clinical endpoints found by phenotypic analyses, the team confirmed roughly 75 percent of the associations described in prior GWAS, and uncovered dozens of SNPs with previously unappreciated ties to phenotypes, including cardiovascular disease-related traits.
In particular, the researchers described nine new associations that reached study-wide significance after the meta-analysis — a set that included associations with everything from severe acne or asthma to high cholesterol. They noted that a missense SNP near the PNPLA3 gene, with known ties to liver traits or conditions, was newly associated with multiple phenotypes such as type 2 diabetes or liver toxicity when taking aspirin or non-steroidal anti-inflammatory drugs.
In a series of follow-up analyses, the team took a closer look at pleiotropic effects present in the PheWAS, and prioritized potential target loci for thromboembolism. For example, the latter analysis highlighted a variant that appeared to reduced the risk of thromboembolisms without increasing the risk of bleeding — a concern for two other associated alleles.
"Whether PheWAS may truly impact decision making during drug development will only become evident with either the emergence of [adverse drug events] in trials that genetics could have predicted, or reduced safety-related attrition rates for portfolios enriched in targets nominated through human genetics," the authors concluded. "The growing number of large-scale population cohorts that link genetic data with extensive health data … will provide opportunities to demonstrate that."