Missing or mis-telling the story? Trade-offs for restriction-site associated compared to whole genome sequencing

Researchers must navigate several trade-offs when deciding which population sequencing method to use. The decision between reduced representation approaches and whole genome sequencing (WGS) impacts marker density, sequencing depth and costs per sample, which will in turn affect the power to accurately characterise certain genomic features, such as regions of the genome exhibiting signals of selection. To investigate the effect of sequencing method on the detection of putatively adaptive regions, we compared selection scan analyses of a set of restriction site-associated DNA sequencing (RADseq) datasets for the common myna (Acridotheres tristis) with a WGS dataset with fewer individuals. Although selection scan statistics were found to be correlated between datasets, no common outliers were found when using outlier thresholds typically applied in such studies. We compared allele frequencies and genotypes across datasets and found that discordances were due to missing markers, different individuals sampled or erroneous genotyping. Most importantly, two regions with strong signals of selection identified through WGS data were missed in the lower density dataset, and population-specific allelic dropout, which can result from restriction enzyme cut site loss in RADseq, created false signals of selection in these datasets. Our results highlight the advantages of WGS over RADseq when used for selection scan analyses, especially for highly structured populations such as those observed in many invasive or endangered species.