Genetic estimates of contemporary effective population size: to what time periods do the estimates apply?

Although most genetic estimates of contemporary effective population size (Ne) are based on models that assume Ne is constant, in real populations Ne changes (often dramatically) over time, and estimates (Ne) will be influenced by Ne in specific generations. In such cases, it is important to properly match Ne to the appropriate time periods (for example, in computing Ne/N ratios). Here I consider this problem for semelparous species with two life histories (discrete generations and variable age at maturity--the 'salmon' model), for two different sampling plans, and for estimators based on single samples (linkage disequilibrium, heterozygote excess) and two samples (temporal method). Results include the following. Discrete generations: (i) Temporal samples from generations 0 and t estimate the harmonic mean Ne in generations 0 through t - 1 but do not provide information about Ne in generation t; (ii) Single samples provide an estimate of Ne in the parental generation, not the generation sampled; (iii) single-sample and temporal estimates never provide information about Ne in exactly the same generations; (iv) Recent bottlenecks can downwardly bias estimates based on linkage disequilibrium for several generations. Salmon model: (i) A pair of single-cohort (typically juvenile) samples from years 0 and t provide a temporal estimate of the harmonic mean of the effective numbers of breeders in the two parental years (N b(0) and N b(t)), but adult samples are more difficult to interpret because they are influenced by Nb in a number of previous years; (ii) For single-cohort samples, both one-sample and temporal methods provide estimates of Nb in the same years (contrast with results for discrete generation model); (iii) Residual linkage disequilibrium associated with past population size will not affect single-sample estimates of Nb as much as in the discrete generation model because the disequilibrium diffuses among different years of breeders. These results lead to some general conclusions about genetic estimates of Ne in iteroparous species with overlapping generations and identify areas in need of further research.     

Comparative estimation of effective population sizes and temporal gene flow in two contrasting population systems

Estimation of effective population sizes (N(e)) and temporal gene flow (N(e)m, m) has many implications for understanding population structure in evolutionary and conservation biology. However, comparative studies that gauge the relative performance of N(e), N(e)m or m methods are few. Using temporal genetic data from two salmonid fish population systems with disparate population structure, we (i) evaluated the congruence in estimates and precision of long- and short-term N(e), N(e)m and m from six methods; (ii) explored the effects of metapopulation structure on N(e) estimation in one system with spatiotemporally linked subpopulations, using three approaches; and (iii) determined to what degree interpopulation gene flow was asymmetric over time. We found that long-term N(e) estimates exceeded short-term N(e) within populations by 2-10 times; the two were correlated in the system with temporally stable structure (Atlantic salmon, Salmo salar) but not in the highly dynamic system (brown trout, Salmo trutta). Four temporal methods yielded short-term N(e) estimates within populations that were strongly correlated, and these were higher but more variable within salmon populations than within trout populations. In trout populations, however, these short-term N(e) estimates were always lower when assuming gene flow than when assuming no gene flow. Linkage disequilibrium data generally yielded short-term N(e) estimates of the same magnitude as temporal methods in both systems, but the two were uncorrelated. Correlations between long- and short-term geneflow estimates were inconsistent between methods, and their relative size varied up to eightfold within systems. While asymmetries in gene flow were common in both systems (58-63% of population-pair comparisons), they were only temporally stable in direction within certain salmon population pairs, suggesting that gene flow between particular populations is often intermittent and/or variable. Exploratory metapopulation N(e) analyses in trout demonstrated both the importance of spatial scale in estimating N(e) and the role of gene flow in maintaining genetic variability within subpopulations. Collectively, our results illustrate the utility of comparatively applying N(e), N(e)m and m to (i) tease apart processes implicated in population structure, (ii) assess the degree of continuity in patterns of connectivity between population pairs and (iii) gauge the relative performance of different approaches, such as the influence of population subdivision and gene flow on N(e) estimation. They further reiterate the importance of temporal sampling replication in population genetics, the value of interpreting N(e)or m in light of species biology, and the need to address long-standing assumptions of current N(e), N(e)m or m models more explicitly in future research.     

Spatial-temporal stratifications in natural populations and how they affect understanding and estimation of effective population size

The concept of effective population size (N(e) ) is based on an elegantly simple idea which, however, rapidly becomes very complex when applied to most real-world situations. In natural populations, spatial and temporal stratifications create different classes of individuals with different vital rates, and this in turn affects (generally reduces) N(e) in complex ways. I consider how these natural stratifications influence our understanding of effective size and how to estimate it, and what the consequences are for conservation and management of natural populations. Important points that emerge include the following: 1 The relative influences of local vs metapopulation N(e) depend on a variety of factors, including the time frame of interest. 2 Levels of diversity in local populations are strongly influenced by even low levels of migration, so these measures are not reliable indicators of local N(e) . 3 For long-term effective size, obtaining a reliable estimate of mutation rate is the most important consideration; unless this is accomplished, estimates can be biased by orders of magnitude. 4 At least some estimators of contemporary N(e) appear to be robust to relatively high (approximately 10%) equilibrium levels of migration, so under many realistic scenarios they might yield reliable estimates of local N(e) . 5 Age structure probably has little effect on long-term estimators of N(e) but can strongly influence contemporary estimates. 6 More research is needed in several key areas: (i) to disentangle effects of selection and drift in metapopulations connected by intermediate levels of migration; (ii) to elucidate the relationship between N(b) (effective number of breeders per year) and N(e) per generation in age-structured populations; (iii) to perform rigorous sensitivity analyses of new likelihood and coalescent-based methods for estimating demographic and evolutionary histories.     

Short-term genetic changes: evaluating effective population size estimates in a comprehensively described brown trout (Salmo trutta) population

The effective population size (N(e)) is notoriously difficult to accurately estimate in wild populations as it is influenced by a number of parameters that are difficult to delineate in natural systems. The different methods that are used to estimate N(e) are affected variously by different processes at the population level, such as the life-history characteristics of the organism, gene flow, and population substructure, as well as by the frequency patterns of genetic markers used and the sampling design. Here, we compare N(e) estimates obtained by different genetic methods and from demographic data and elucidate how the estimates are affected by various factors in an exhaustively sampled and comprehensively described natural brown trout (Salmo trutta) system. In general, the methods yielded rather congruent estimates, and we ascribe that to the adequate genotyping and exhaustive sampling. Effects of violating the assumptions of the different methods were nevertheless apparent. In accordance with theoretical studies, skewed allele frequencies would underestimate temporal allele frequency changes and thereby upwardly bias N(e) if not accounted for. Overlapping generations and iteroparity would also upwardly bias N(e) when applied to temporal samples taken over short time spans. Gene flow from a genetically not very dissimilar source population decreases temporal allele frequency changes and thereby acts to increase estimates of N(e). Our study reiterates the importance of adequate sampling, quantification of life-history parameters and gene flow, and incorporating these data into the N(e) estimation.     

Monitoring the effective population size of a brown bear (Ursus arctos) population using new single-sample approaches

The effective population size (N(e) ) could be the ideal parameter for monitoring populations of conservation concern as it conveniently summarizes both the evolutionary potential of the population and its sensitivity to genetic stochasticity. However, tracing its change through time is difficult in natural populations. We applied four new methods for estimating N(e) from a single sample of genotypes to trace temporal change in N(e) for bears in the Northern Dinaric Mountains. We genotyped 510 bears using 20 microsatellite loci and determined their age. The samples were organized into cohorts with regard to the year when the animals were born and yearly samples with age categories for every year when they were alive. We used the Estimator by Parentage Assignment (EPA) to directly estimate both N(e) and generation interval for each yearly sample. For cohorts, we estimated the effective number of breeders (N(b) ) using linkage disequilibrium, sibship assignment and approximate Bayesian computation methods and extrapolated these estimates to N(e) using the generation interval. The N(e) estimate by EPA is 276 (183-350 95% CI), meeting the inbreeding-avoidance criterion of N(e) > 50 but short of the long-term minimum viable population goal of N(e) > 500. The results obtained by the other methods are highly consistent with this result, and all indicate a rapid increase in N(e) probably in the late 1990s and early 2000s. The new single-sample approaches to the estimation of N(e) provide efficient means for including N(e) in monitoring frameworks and will be of great importance for future management and conservation.     

Can Genetic Estimators Provide Robust Estimates of the Effective Number of Breeders in Small Populations?

The effective population size (N_e) is proportional to the loss of genetic diversity and the rate of inbreeding, and its accurate estimation is crucial for the monitoring of small populations. Here, we integrate temporal studies of the gecko Oedura reticulata, to compare genetic and demographic estimators of N_e. Because geckos have overlapping generations, our goal was to demographically estimate N_bI, the inbreeding effective number of breeders and to calculate the N_bI/N_a ratio (N_a = number of adults) for four populations. Demographically estimated N_bI ranged from 1 to 65 individuals. The mean reduction in the effective number of breeders relative to census size (N_bI/N_a) was 0.1 to 1.1. We identified the variance in reproductive success as the most important variable contributing to reduction of this ratio. We used four methods to estimate the genetic based inbreeding effective number of breeders N_bI(gen) and the variance effective populations size N_eV(gen) estimates from the genotype data. Two of these methods - a temporal moment-based (MBT) and a likelihood-based approach (TM3) require at least two samples in time, while the other two were single-sample estimators - the linkage disequilibrium method with bias correction LDNe and the program ONeSAMP. The genetic based estimates were fairly similar across methods and also similar to the demographic estimates excluding those estimates, in which upper confidence interval boundaries were uninformative. For example, LDNe and ONeSAMP estimates ranged from 14–55 and 24–48 individuals, respectively. However, temporal methods suffered from a large variation in confidence intervals and concerns about the prior information. We conclude that the single-sample estimators are an acceptable short-cut to estimate N_bI for species such as geckos and will be of great importance for the monitoring of species in fragmented landscapes.     

Comparative analyses of effective population size within and among species: ranid frogs as a case study

It has recently become practicable to estimate the effective sizes (N(e) ) of multiple populations within species. Such efforts are valuable for estimating N(e) in evolutionary modeling and conservation planning. We used microsatellite loci to estimate N(e) of 90 populations of four ranid frog species (20-26 populations per species, mean n per population = 29). Our objectives were to determine typical values of N(e) for populations of each species, compare N(e) estimates among the species, and test for correlations between several geographic variables and N(e) within species. We used single-sample linkage disequilibrium (LD), approximate Bayesian computation (ABC), and sibship assignment (SA) methods to estimate contemporary N(e) for each population. Three of the species-Rana pretiosa, R. luteiventris, and R. cascadae- have consistently small effective population sizes (<50). N(e) in Lithobates pipiens spans a wider range, with some values in the hundreds or thousands. There is a strong east-to-west trend of decreasing N(e) in L. pipiens. The smaller effective sizes of western populations of this species may be related to habitat fragmentation and population bottlenecking.     

Effect of unsampled populations on the estimation of population sizes and migration rates between sampled populations

Current estimators of gene flow come in two methods; those that estimate parameters assuming that the populations investigated are a small random sample of a large number of populations and those that assume that all populations were sampled. Maximum likelihood or Bayesian approaches that estimate the migration rates and population sizes directly using coalescent theory can easily accommodate datasets that contain a population that has no data, a so-called 'ghost' population. This manipulation allows us to explore the effects of missing populations on the estimation of population sizes and migration rates between two specific populations. The biases of the inferred population parameters depend on the magnitude of the migration rate from the unknown populations. The effects on the population sizes are larger than the effects on the migration rates. The more immigrants from the unknown populations that are arriving in the sample populations the larger the estimated population sizes. Taking into account a ghost population improves or at least does not harm the estimation of population sizes. Estimates of the scaled migration rate M (migration rate per generation divided by the mutation rate per generation) are fairly robust as long as migration rates from the unknown populations are not huge. The inclusion of a ghost population does not improve the estimation of the migration rate M; when the migration rates are estimated as the number of immigrants Nm then a ghost population improves the estimates because of its effect on population size estimation. It seems that for 'real world' analyses one should carefully choose which populations to sample, but there is no need to sample every population in the neighbourhood of a population of interest.     

Historical Analysis of Genetic Variation Reveals Low Effective Population Size in a Northern Pike (Esox Lucius) Population

Effective population size (N(e)) of a natural fish population was estimated from temporal changes in allele frequencies at seven microsatellite loci. Use of a historical collection of fish scales made it possible to increase the precision of estimates by increasing the time interval between samples and to use an equation developed for discrete generations without correcting for demographic parameters. Estimates of N(e) for the time intervals 1961-1977 and 1977-1993 were 35 and 72, respectively. For the entire interval, 1961-1993, the estimate of N(e) was 48 when based on a weighted mean derived from the above two estimates or 125 when calculated from 1961 and 1993 samples only. Corresponding ratios of effective size to adult census size ranged from 0.03 to 0.14. An N(e) of 48 over a 32-year period would imply that this population lost as much as 8% of its heterozygosity in that time. Results suggest the potential for using genetic methods based on microsatellite loci data to compare historical trends in N(e) with population dynamic parameters. Such comparisons will help to evaluate the relationship between genetic diversity and long-term persistence of natural populations.     

Estimating effective population size and migration rates from genetic samples over space and time

In the past, moment and likelihood methods have been developed to estimate the effective population size (N(e)) on the basis of the observed changes of marker allele frequencies over time, and these have been applied to a large variety of species and populations. Such methods invariably make the critical assumption of a single isolated population receiving no immigrants over the study interval. For most populations in the real world, however, migration is not negligible and can substantially bias estimates of N(e) if it is not accounted for. Here we extend previous moment and maximum-likelihood methods to allow the joint estimation of N(e) and migration rate (m) using genetic samples over space and time. It is shown that, compared to genetic drift acting alone, migration results in changes in allele frequency that are greater in the short term and smaller in the long term, leading to under- and overestimation of N(e), respectively, if it is ignored. Extensive simulations are run to evaluate the newly developed moment and likelihood methods, which yield generally satisfactory estimates of both N(e) and m for populations with widely different effective sizes and migration rates and patterns, given a reasonably large sample size and number of markers.     

Temporal estimates of effective population size in species with overlapping generations

The standard temporal method for estimating effective population size (N(e)) assumes that generations are discrete, but it is routinely applied to species with overlapping generations. We evaluated bias in the estimates N(e) caused by violation of this assumption, using simulated data for three model species: humans (type I survival), sparrow (type II), and barnacle (type III). We verify a previous proposal by Felsenstein that weighting individuals by reproductive value is the correct way to calculate parametric population allele frequencies, in which case the rate of change in age-structured populations conforms to that predicted by discrete-generation models. When the standard temporal method is applied to age-structured species, typical sampling regimes (sampling only newborns or adults; randomly sampling the entire population) do not yield properly weighted allele frequencies and result in biased N(e). The direction and magnitude of the bias are shown to depend on the sampling method and the species' life history. Results for populations that grow (or decline) at a constant rate paralleled those for populations of constant size. If sufficient demographic data are available and certain sampling restrictions are met, the Jorde-Ryman modification of the temporal method can be applied to any species with overlapping generations. Alternatively, spacing the temporal samples many generations apart maximizes the drift signal compared to sampling biases associated with age structure.     

Estimation of migration rates from marker‐based parentage analysis

Coupled with rapid developments of efficient genetic markers, powerful population genetic methods were proposed to estimate migration rates (m) in natural populations in much broader spatial and temporal scales than the traditional mark‐release‐recapture (MRR) methods. Highly polymorphic (e.g. microsatellites) and genomic‐wide (e.g. SNPs) markers provide sufficient information to assign individuals to their populations or parents of origin and thereby to estimate directly m in a way similar to MRR. Such direct estimates of current migration rates are particularly useful in understanding the ecology and microevolution of wild populations and in managing the populations in the future. In this study, I proposed and implemented, in the software MigEst, a likelihood method to use marker‐based parentage assignments in jointly estimating m and candidate parent sampling proportions (x) in a subset of populations, investigated its power and accuracy using data simulated in various scenarios of population properties (e.g. the actual m, number, size and differentiation of populations) and sampling properties (e.g. the numbers of sampled parent candidates, offspring and markers), compared it with the population assignment approach implemented in the software BayesAss and demonstrated its usefulness by analysing a microsatellite data set from three natural populations of Brazilian bats. Simulations showed that MigEst provides unbiased and accurate estimates of m and performs better than BayesAss except when populations are highly differentiated with very small and ecologically insignificant migration rates. A valuable property of MigEst is that in the presence of unsampled populations, it gives good estimates of the rate of migration among sampled populations as well as of the rate of migration into each sampled population from the pooled unsampled populations.     

Joint estimation of migration rate and effective population size using the island model

Using the island model of population demography, I report that the demographic parameters migration rate and effective population size can be jointly estimated with equilibrium probabilities of identity in state calculated using a sample of genotypes collected at a single point in time from a single generation. The method, which uses moment-type estimators, applies to dioecious populations in which females and males have identical demography and monoecious populations with no selfing and requires that offspring genotypes are sampled following reproduction and prior to migration. I illustrate the estimation procedure using the infinite-island model with no mutation and the finite-island model with three kinds of mutation models. In the infinite-island model with no mutation, the estimators can be expressed as simple functions of estimates of the F-statistic parameters F(IT) and F(ST). In the finite-island model with mutation among k alleles, mutation rate, migration rate, and effective population size can be simultaneously estimated. The estimates of migration rate and effective population size are somewhat robust to violations in assumptions that may arise in empirical applications such as different kinds of mutation models and deviations from temporal equilibrium.     

Genetic monitoring reveals temporal stability over 30 years in a small, lake-resident brown trout population

Knowledge of the degree of temporal stability of population genetic structure and composition is important for understanding microevolutionary processes and addressing issues of human impact of natural populations. We know little about how representative single samples in time are to reflect population genetic constitution, and we explore the temporal genetic variability patterns over a 30-year period of annual sampling of a lake-resident brown trout (Salmo trutta) population, covering 37 consecutive cohorts and five generations. Levels of variation remain largely stable over this period, with no indication of substructuring within the lake. We detect genetic drift, however, and the genetically effective population size (N(e)) was assessed from allele-frequency shifts between consecutive cohorts using an unbiased estimator that accounts for the effect of overlapping generation. The overall mean N(e) is estimated as 74. We find indications that N(e) varies over time, but there is no obvious temporal trend. We also estimated N(e) using a one-sample approach based on linkage disequilibrium (LD) that does not account for the effect of overlapping generations. Combining one-sample estimates for all years gives an N(e) estimate of 76. This similarity between estimates may be coincidental or reflecting a general robustness of the LD approach to violations of the discrete generations assumption. In contrast to the observed genetic stability, body size and catch per effort have increased over the study period. Estimates of annual effective number of breeders (N(b)) correlated with catch per effort, suggesting that genetic monitoring can be used for detecting fluctuations in abundance.     

Long-term stability and effective population size in North Sea and Baltic Sea cod (Gadus morhua)

DNA from archived otoliths was used to explore the temporal stability of the genetic composition of two cod populations, the Moray Firth (North Sea) sampled in 1965 and 2002, and the Bornholm Basin (Baltic Sea) sampled in 1928 and 1997. We found no significant changes in the allele frequencies for the Moray Firth population, while subtle but significant genetic changes over time were detected for the Bornholm Basin population. Estimates of the effective population size ( N _e ) generally exceeded 500 for both populations when employing a number of varieties of the temporal genetic method. However, confidence intervals were very wide and N _e 's most likely range in the thousands. There was no apparent loss of genetic variability and no evidence of a genetic bottleneck for either of the populations. Calculations of the expected levels of genetic variability under different scenarios of N _e showed that the number of alleles commonly reported at microsatellite loci in Atlantic cod is best explained by N _e 's exceeding thousand. Recent fishery-induced bottlenecks can, however, not be ruled out as an explanation for the apparent discrepancy between high levels of variability and recently reported estimates of N _e  << 1000. From life history traits and estimates of survival rates in the wild, we evaluate the compatibility of the species' biology and extremely low N _e / N ratios. Our data suggest that very small N _e 's are not likely to be of general concern for cod populations and, accordingly, most populations do not face any severe threat of losing evolutionary potential due to genetic drift.     

A potential bias in the temporal method for estimating Ne in admixed populations under natural selection

Indirect genetic methods are frequently used to estimate the effective population size (N(e)) or effective number of breeders (N(b)) in natural populations. Although assumptions behind these methods are often violated, there have been few attempts to evaluate how accurate these estimates really are in practice. Here we investigate the influence of natural selection following a population admixture on the temporal method for estimating N(e). Our analytical and simulation results suggest that N(e) is often underestimated in this method when subpopulations differ substantially in allele frequencies and in reproductive success. The underestimation is exacerbated when true N(e) and the fraction of the low-fitness group are large. As an empirical example, we compared N(b) estimated in natural populations of steelhead trout (Oncorhynchus mykiss) using the temporal method (N(b[temp])) with estimates based on direct demographic methods (N(b[demo])) and the linkage disequilibrium method (N(b[LD])). While N(b[LD]) was generally in close agreement with N(b[demo]), N(b[temp]) was much lower in sample sets that were dominated by nonlocal hatchery fish with low reproductive success, as predicted by the analytical results. This bias in the temporal method, which arises when genes associated with a particular group of parents are selected against in the offspring sample, has not been widely appreciated before. Such situations may be particularly common when artificial propagation or translocations are used for conservation. The linkage disequilibrium method, which requires data from only one sample, is robust to this type of bias, although it can be affected by other factors.     

Demographic and genetic estimates of effective population size (Ne) reveals genetic compensation in steelhead trout

Estimates of effective population size (Ne) are required to predict the impacts of genetic drift and inbreeding on the evolutionary dynamics of populations. How the ratio of Ne to the number of sexually mature adults (N) varies in natural vertebrate populations has not been addressed. We examined the sensitivity of Ne/N to fluctuations of N and determined the major variables responsible for changing the ratio over a period of 17 years in a population of steelhead trout (Oncorhynchus mykiss) from Washington State. Demographic and genetic methods were used to estimate Ne. Genetic estimates of Ne were gained via temporal and linkage disequilibrium methods using data from eight microsatellite loci. DNA for genetic analysis was amplified from archived smolt scales. The Ne/N from 1977 to 1994, estimated using the temporal method, was 0.73 and the comprehensive demographic estimate of Ne/N over the same time period was 0.53. Demographic estimates of Ne indicated that variance in reproductive success had the most substantial impact on reducing Ne in this population, followed by fluctuations in population size. We found increased Ne/N ratios at low N, which we identified as genetic compensation. Combining the information from the demographic and genetic methods of estimating Ne allowed us to determine that a reduction in variance in reproductive success must be responsible for this compensation effect. Understanding genetic compensation in natural populations will be valuable for predicting the effects of changes in N (i.e. periods of high population density and bottlenecks) on the fitness and genetic variation of natural populations.     

Concordant genetic estimators of migration reveal anthropogenically enhanced source-sink population structure in the river sculpin, Cottus gobio

River systems are vulnerable to natural and anthropogenic habitat fragmentation and will often harbor populations deviating markedly from simplified theoretical models. We investigated fine-scale population structure in the sedentary river fish Cottus gobio using microsatellites and compared migration estimates from three FST estimators, a coalescent maximum-likelihood method and Bayesian recent migration analyses. Source-sink structure was evident via asymmetry in migration and genetic diversity with smaller upstream locations emigration biased and larger downstream subpopulations immigration biased. Patterns of isolation by distance suggested that the system was largely, but not entirely, in migration-drift equilibrium, with headwater populations harboring a signal of past colonizations and in some cases also recent population bottlenecks. Up- vs. downstream asymmetry in population structure was partly attributable to the effects of flow direction, but was enhanced by weirs prohibiting compensatory upstream migration. Estimators of migration showed strong correspondence, at least in relative terms, especially if pairwise FST was used as an indirect index of relative gene flow rather than being translated to Nm. Since true parameter values are unknown in natural systems, comparisons among estimators are important, both to determine confidence in estimates of migration and to validate the performance of different methods.     

GONe: software for estimating effective population size in species with generational overlap

GONe is a user-friendly, Windows-based program for estimating effective size (N(e) ) in populations with overlapping generations. It uses the Jorde-Ryman modification to the temporal method to account for age structure in populations. This method requires estimates of age-specific survival and birth rate and allele frequencies measured in two or more consecutive cohorts. Allele frequencies are acquired by reading in genotypic data from files formatted for either GENEPOP or TEMPOFS. For each interval between consecutive cohorts, N(e) is estimated at each locus and over all loci. Furthermore, N(e) estimates are output for three different genetic drift estimators (F(s) , F(c) and F(k) ). Confidence intervals are derived from a chi-square distribution with degrees of freedom equal to the number of independent alleles. GONe has been validated over a wide range of N(e) values, and for scenarios where survival and birth rates differ between sexes, sex ratios are unequal and reproductive variances differ. GONe is freely available for download at https://bcrc.bio.umass.edu/pedigreesoftware/.     

Comparative evaluation of a new effective population size estimator based on approximate bayesian computation

We describe and evaluate a new estimator of the effective population size (N(e)), a critical parameter in evolutionary and conservation biology. This new "SummStat" N(e) estimator is based upon the use of summary statistics in an approximate Bayesian computation framework to infer N(e). Simulations of a Wright-Fisher population with known N(e) show that the SummStat estimator is useful across a realistic range of individuals and loci sampled, generations between samples, and N(e) values. We also address the paucity of information about the relative performance of N(e) estimators by comparing the SummStat estimator to two recently developed likelihood-based estimators and a traditional moment-based estimator. The SummStat estimator is the least biased of the four estimators compared. In 32 of 36 parameter combinations investigated using initial allele frequencies drawn from a Dirichlet distribution, it has the lowest bias. The relative mean square error (RMSE) of the SummStat estimator was generally intermediate to the others. All of the estimators had RMSE > 1 when small samples (n = 20, five loci) were collected a generation apart. In contrast, when samples were separated by three or more generations and N(e) < or = 50, the SummStat and likelihood-based estimators all had greatly reduced RMSE. Under the conditions simulated, SummStat confidence intervals were more conservative than the likelihood-based estimators and more likely to include true N(e). The greatest strength of the SummStat estimator is its flexible structure. This flexibility allows it to incorporate any potentially informative summary statistic from population genetic data.