Estimating current effective sizes of large populations from a single sample of genomic marker data: A comparison of estimators by simulations

Abstract Genome‐wide single nucleotide polymorphisms (SNPs) data are increasingly used in estimating the current effective population sizes ( N e ) for informing the conservation of endangered species and guiding the management of exploited species. Previous assessments of sibship frequency (SF) and linkage disequilibrium (LD) estimators of N e focused on small populations where genetic drift is strong and thus N e is easy to estimate. Genomic single nucleotide polymorphism (SNP) data provide ample information and hold the potential for application of these estimators to large populations where genetic drift is rather weak and thus N e is difficult to estimate. In this study, I simulated very large populations and sampled a widely variable number of individuals (genotyped at 10,000 SNPs) for estimating N e by both SF and LD methods. I also considered the more realistic situation where a population experiences a bottleneck, and where marker data suffer from genotyping errors. The simulations show that both SF and LD methods can yield accurate N e estimates of very large populations when sampled individuals are sufficiently numerous. When n is much smaller than N e , however, N e estimates are in a bimodal distribution with a substantial proportion of the estimates being infinitely large. For a population with a bottleneck, LD estimator overestimates and underestimates the N e of the parental population from samples taken at and after the bottleneck, respectively. LD estimator also overestimates N e substantially when applied to data suffering from allelic dropouts and false alleles. In contrast, SF estimator is unbiased and accurate when populations are changing in size or markers suffer from genotyping errors.