Evaluation of the minimum sampling design for population genomic and microsatellite studies. An analysis based on wild maize.

Massive parallel sequencing is revolutionizing the field of molecular ecology by allowing to understand better the evolutionary history of populations and species, and to detect genomic regions that could be under selection. However, the needed economic and computational resources generate a tradeoff between the amount of loci that can be obtained and the number of populations or individuals that can be sequenced. In this work, we analyzed and compared two extensive genomic and one large microsatellite datasets consisting of empirical data. We generated different subsampling designs by changing the number of loci, individuals, populations and individuals per population to test for deviations in classic population genetics parameters (HS, FIS, FST) and landscape genetic tests (isolation by distance and environment, central abundance hypothesis). We also tested the effect of sampling different number of populations in the detection of outlier SNPs. We found that the microsatellite dataset is very sensitive to the number of individuals sampled when obtaining summary statistics. FIS was particularly sensitive to a low sampling of individuals in the genomic and microsatellite datasets. For the genomic datasets, we found that as long as many populations are sampled, few individuals and loci are needed. For all datasets we found that increasing the number of population sampled is important to obtain precise landscape genetic estimates. Finally, we corroborated that outlier tests are sensitive to the number of populations sampled. We conclude by proposing different sampling designs depending on the objectives.