While PheWAS has great potential, this approach faces several hurdles to more widespread application. The first is simply the technical and logistical challenge of generating a phenome. Intense collaborative efforts are a sine qua non even for the most tractable model organisms such as Drosophila36. The second is yet another example of the multiple testing problem: what is the appropriate correction given the size of the phenome and its structure? We have computed false discovery rate (FDR) q values at a conservative threshold and have aligned our results, when possible, with the BioVU clinical cohort. However, in both species, the selection of appropriate q values will depend on the purpose of studies and the relative costs of types 1 and 2 errors. Effective solutions may require adjusting thresholds based on the scope and intent of studies, as well as prior information about gene-to-phenotype relations. Alignment of phenotype associations across both humans and mice, however, adds validity to both. Very large, densely genotyped or sequenced populations will be needed to more deeply interrogate the human phenome. The third problem is linkage
