Effects of sample size on various genetic diversity measures in population genetic study with microsatellite DNA markers

摘要

Sample collection and sampling strategy is of primary importance in population genetic studies, and can greatly influence the analysis and interpretation of the data obtained. Microsatellite DNA sequences are the most revealing DNA markers available so far for inferring population structure and dynamics, and have been widely employed in genetic studies of populations. Nevertheless, few studies have specifically examined the effects of sample size and other sampling factors on various genetic diversity measures in population genetic work using these markers. When such issues were discussed in the literature, it was often very brief or not based on experimental data. Here, we examined these issues more closely using empirical genotype data from an analysis of 26 populations of the migratory locusts Locusta migratoria (1 381 individuals) at 8 microsatellite loci. The following genetic diversity measures were studied: number of alleles per locus (NA), mean number of allele over loci (MNA), observed heterozygosity (Ho) and expected heterozygosity (He). We also tried to infer the theoretic minimum sample size needed for population genetic studies with microsatellite loci. Our main results and conclusions are as follows. （ 1 ） NA and MNA, i.e. the estimated allelic richness of populations, are significantly affected by sample size (see Figs 1 - 4, Table 2), while no significant correlation between sample size and heterozygosities (He and Ho) was observed (Figs 6 - 7). （ 2 ） Both the rarefaction method and random resampling method can remove effects of sample size on the allelic richness measures (NA and MNA) discussed above (Fig. 5, Figs 8 - 9). Therefore, although the uncorrected NA and MNA are sensitive to sample size variation and should not be reliable parameters for assessing demographic processes such as bottleneck events, these measures after correction can be used to infer historical population size reduction. Measures of heterogygosity (Ho and He) appears to be fairly stable and suitable for studying genic variation of populations. （ 3 ） The extent to which NA and MNA were affected by sample size varies among microsatellite loci. Sample size variation has much greater impact on highly polymorphic loci ( e.g. LmIOZc19 locus of the migratory locust) than on less polymorphic loci ( e.g. LmIOZc76 locus; see Figs 1 and 3). （ 4 ） The minimum sample size needed in a study depends much on the objective of the research project, and is determined by the population frequency of the least represented allele (s). In order to detect an allele of the frequency 0.01 in population with 95% probability, 149 or more samples are required (Table 3). Given that most microsatellite DNA loci have 20 to 30 alleles, at least 30-50 individuals should be analyzed in most cases to detect all these alleles. This theoretical value, calculated from the formula P₀ = (1 － q ) ^{2 n} (where n is the sample size, q is the frequency of the allele in question, P₀ is the probability of the allele being not represented in the sample), agrees well with simulation results (see Fig.3). （ 5 ） Finally, we strongly advise that homogeneity tests should be carried out between samples collected at different times from the same location or between sexes before pooling such samples, to check potential heterogeneity due to temporal variation in population structure or sex-bias dispersal. Although with the migratory locusts we studied, no significant difference exists among such samples, this may not be the case for other study systems [ Acta Zoologica Sinica 50 (2): 279 - 290 , 2004].