Accurate gene diversity estimates from amplified fragment length polymorphism (AFLP) markers

Three procedures for the estimation of null allele frequencies and gene diversity from dominant multilocus data were empirically tested in natural populations of the outcrossing angiosperm Persoonia mollis (Proteaceae). The three procedures were the square root transform of the null homozygote frequency, the Lynch & Milligan procedure, and the Bayesian method. Genotypes for each of 116 polymorphic loci generated by amplified fragment length polymorphism (AFLP) were inferred from segregation patterns in progeny arrays. Therefore, for the plus phenotype (band present), heterozygotes were distinguished from homozygotes. In contrast to previous studies, all three procedures produced very similar mean estimates of heterozygosity, which were in turn accurate estimators of the direct value (HO = 0.28). A second population of P. mollis displayed markedly lower levels of heterozygosity (HO = 0.20) but approximately twice as many polymorphic loci (284). These AFLP results show that biases in estimates of average null allele frequency and heterozygosity are largely eliminated in highly polymorphic dominant marker data sets displaying a J-shaped beta distribution with a high percentage of loci containing more than three null homozygotes and relatively few loci with no null homozygotes. This distribution may be typical of outcrossing angiosperms.