Temporal-method estimates of N_e from highly polymorphic loci

摘要

The temporal method is used widely to estimate genetic effective population size (N_e), a parameter of fundamental interest to studies of evolutionary and conservation biology. The statistical properties of temporal-method estimates have not been explored for highly polymorphic DNA markers that often contain many alleles occurring in very low frequencies. We used a Monte Carlo simulation approach to assess accuracy and precision of the temporal method when implemented with haplotypic/allelic data at mitochondrial (mt)DNA and nuclear-encoded microsatellite DNA loci. Estimates of N_e were between 2%-106% greater than their true values in 48 simulations parameterized using different demographic scenarios, models of mutation, and sample sizes. Overestimation of N_e results from a bias in the approximation used by Waples (1989) to derive the relationship between the expected temporal variance (F) and N_e when allele frequencies are very close to 0 or 1. Our results show that one commonly applied solution to this problem, binning of low-frequency alleles, results in a trade-off of accuracy and precision in some cases. We show that both chi-square and normal approximations are appropriate for estimating 95% confidence intervals of N_e and we develop a power analysis based on the chi-square distribution to estimate sample sizes and allelic diversity required to evaluate specific hypotheses. For highly polymorphic loci like mtDNA and microsatellites, the increased precision afforded by the presence of rare alleles outweighs the upward bias in temporal-method estimates of N_e.