Optimizing strain selection for association studies under hard cost constraints

Quantitative genetics methods linking genotype to phenotype are powerful tools in model organisms and nonhuman populations, allowing researchers to apply well-controlled perturbations to commercially available strain collections. However, purchasing and phenotyping large collections can be cost-prohibitive, and it is unclear how to select smaller subsets to maximize statistical power for a given budget. We evaluated several strain selection methods considering cost, genetic diversity, or both, using simulations from 2 settings-phylogenetic regression across bacterial isolates and linear mixed-model regression across mouse strains-as well as real subsampled data on cardiovascular phenotypes from the Hybrid Mouse Diversity Panel. Surprisingly, considering only costs while ignoring genetic diversity (MinCost) was typically one of the most powerful approaches. Considering diversity alone tended to yield low power, likely because it resulted in fewer total strains being selected. Methods weighting both cost and diversity could sometimes outperform MinCost, but this depended on the scenario and how diversity was maximized. Counterintuitively, the most commonly studied diversity-maximizing objective-maximizing the minimum pairwise distance between selected strains (MaxMin)-had the lowest power. The most powerful objectives included 2 not previously applied for this purpose: maximizing the total sum of pairwise genetic distances (MaxSum) and maximizing total minor allele frequency (MaxMAF). A new cost-aware p-Median approach, which selects the most representative strain subset, also performed well, especially on real data. Overall, when strains have unequal costs, maximizing sample size under a given budget should typically take priority over maximizing genetic diversity, though some methods co-optimizing both may retain more power.