Detecting Quantitative Trait Loci via Experimental Evolution: an Analysis of Power.

A novel approach for dissecting complex traits is to experimentally evolve laboratory populations under a controlled environment shift, resequence the resulting populations, and identify single nucleotide polymorphisms (SNPs) and/or genomic regions highly diverged in allele frequency. This experimental design seems promising, but the power to detect these loci is not deeply understood. To better understand the power and localization ability of such an evolve and resequence (E&R) approach, we carried out forward-in-time population genetics simulations of 1 Mb genomic regions under a large combination of experimental conditions, then attempted to detect significantly diverged SNPs. Our analysis indicates that the ability to detect differentiation between populations is primarily affected by selection coefficient, population size, number of replicate populations, and number of founding haplotypes. We estimate that E&R studies can detect and localize causative sites with above 80% success when the number of founder haplotypes is over 500, experimental populations are replicated 25-fold minimum, population size is at least 1,000 diploid individuals, and the selection coefficient on the locus of interest is at least 0.1. More achievable experimental designs (less replicated, fewer founder haplotypes, smaller effective population size, and smaller selection coefficients) can have power of greater than 50% to identify a handful of SNPs of which one is likely causative. In cases where the selection coefficient is above 0.2, less demanding experimental designs can yield high power. Other groups have had similar results to these. We also present evidence that large numbers of causative sites across the genome decrease localization power.