Estimating Contemporary Effective Population Size on the Basis of Linkage Disequilibrium in the Face of Migration

Authors

Robin Waples, NOAAFollow
Phillip R. England, CSIRO Marine and Atmospheric Research and Wealth from Oceans FlagshipFollow

Date of this Version

10-2011

Citation

Genetics, Vol. 189, 633–644 October 2011

Abstract

Effective population size (N_e) is an important genetic parameter because of its relationship to loss of genetic variation, increases in inbreeding, accumulation of mutations, and effectiveness of selection. Like most other genetic approaches that estimate contemporary N_e, the method based on linkage disequilibrium (LD) assumes a closed population and (in the most common applications) randomly recombining loci. We used analytical and numerical methods to evaluate the absolute and relative consequences of two potential violations of the closed-population assumption: (1) mixture LD caused by occurrence of more than one gene pool, which would downwardly bias ^{^}N_e, and (2) reductions in drift LD (and hence upward bias in ^{^}N_e) caused by an increase in the number of parents responsible for local samples. The LD method is surprisingly robust to equilibrium migration. Effects of mixture LD are small for all values of migration rate (m), and effects of additional parents are also small unless m is high in genetic terms. LD estimates of N_e therefore accurately reflect local (subpopulation) N_e unless m > ~5–10%. With higher m, ^{^}N_e converges on the global (metapopulation) N_e. Two general exceptions were observed. First, equilibrium migration that is rare and hence episodic can occasionally lead to substantial mixture LD, especially when sample size is small. Second, nonequilibrium, pulse migration of strongly divergent individuals can also create strong mixture LD and depress estimates of local N_e. In both cases, assignment tests, Bayesian clustering, and other methods often will allow identification of recent immigrants that strongly influence results. In simulations involving equilibrium migration, the standard LD method performed better than a method designed to jointly estimate Ne and m. The above results assume loci are not physically linked; for tightly linked loci, the LD signal from past migration events can persist for many generations, with consequences for N_e estimates that remain to be evaluated.