The Illusion of Polygenicity in Poolseq studies: Insufficient Power Can Mask Simple Genetic Architectures

In Drosophila melanogaster, genome-wide association studies (GWAS) and quantitative trait locus (QTL) mapping have often yielded seemingly contradictory results, with GWAS typically suggesting highly polygenic architectures while QTL mapping identifies fewer loci of larger effect. We investigated this discrepancy using a published mapping study that contrasted pools of control flies with pools of flies representing the most zinc chloride-resistant larvae. The dataset provided a unique opportunity to compare three distinct analytical approaches: an X-QTL mapping strategy based on inferring known founder haplotype frequencies, a SNP-by-SNP GWAS using directly ascertained SNP counts, and a SNP-by-SNP GWAS using imputed SNP counts derived from estimated haplotype frequencies. Our analysis revealed that X-QTL and imputed SNP approaches uncovered a relatively simple genetic architecture dominated by several major-effect loci. In contrast, the traditional SNP-by-SNP GWAS suggested a highly polygenic trait with numerous small-effect loci dispersed throughout the genome. We demonstrate that this apparent discrepancy primarily stems from the limited statistical power of the SNP-by-SNP GWAS approach when applied to pooled sequencing data. Despite ~700X sequence coverage and greater than 3000 flies per treatment, the inferred polygenic architecture under the SNP-by-SNP approach appears to simply be an artifact of insufficient power to detect major effect loci. Through simulations, we illustrate that reliably detecting subtle allele frequency differences between pools requires substantially larger sample sizes and higher sequencing coverage than typically employed in current studies. Our findings highlight the importance of considering statistical power and methodological approaches when interpreting genetic architecture studies.