Genotyping errors
Genotyping errors were quantified by comparing original AFLP profiles to replicates for the 47 individuals replicated. Mean error rate per phenotype was calculated following Pompanon et al.(2005). In addition, mean error rate per fragment amplified was defined as ef = mf / n, where n is the number of loci x individuals for which at least one of the two replicates had a band, and mf is the fraction of n that presented a mismatch between replicates. Thus, ef assumes that genotyping errors are essentially due to fragment dropouts and is less conservative because n does not include recessive phenotypes without mismatches. This assumption appears plausible since false bands at a truly recessive phenotype position must be rare given the perfect adaptor/primer match and the selectivity of the 3’-end of the AFLP primer.
The mean error rate per AFLP phenotype was 0.85%, a value within the range reported in other studies (Pompanon et al. 2005). However, two samples accounted for 46% of all errors. Clearly, the replicate for these samples did not amplify well and were thus excluded from subsequent analyses. Error rate per phenotype without these two samples was 0.48%. Distribution of errors across loci was close to uniformity (Kolmogorov-Smirnov goodness-of-fit : d = 7, n = 32 loci with errors, P = 0.05) but not across samples (d = 9, n = 18 samples with errors, P < 0.001), when considering loci or samples with at least one genotypic mismatch. Hence, eliminating problematic individuals rather than loci should better improve the accuracy of our results. In fact, whenever an individual was suspected to exhibit errors because of an “unusual” genetic dissimilarity with other birds, we did re-amplify for verification. As predicted, error rate per amplified fragment was higher (ef = 0.59) than error rate per phenotype (i.e., per locus x sample) but only slightly so. Overall, error rates are low and likely to have a negligible effect on results. For example, for a fixed (monomorphic) fragment to become erroneously polymorphic under the 5%-criterion, it would necessitate a fragment dropout in at least 20 wandering albatrosses or 1.7 (≈2) Amsterdam albatrosses, while our error rate predicts dropouts for 2 and 0.17 birds, for the two species respectively, assuming uniform distribution of errors across loci.
Homoplasy
Homoplasy, i.e. the occurrence of non-homologous loci of identical size, may also bias estimates of genetic diversity. Vekemans et al. (2002) showed that the AFLP technique generates a non-uniform distribution of fragment sizes, such that the proportion of detectable fragments greatly decreases with size. This, in turn, leads to a greater level of homoplasy for small fragments. As a consequence, the proportion of polymorphic fragments increases as a function of size (Vekemans et al. 2002). In order to assess whether polymorphism varies with fragment density/size, we calculated the proportion of polymorphic loci (P5%) separately for each 50 bp interval between 50 and 300 bp. Fragments above > 300 bp were rare and thus pooled. Larger fragments (150-300 bp) were rarer than shorter fragments (50-150 bp; χ2 = 15.71, P < 0.01, df = 5; Table 1). They also showed a higher level of polymorphism, but this result was not significant (d = 3.8, n = 12 polymorphic loci, P > 0.2). The density of scored markers was very low for CG-primers (respectively 1.0 locus/10 bp and 0.9 loci/10 bp of AFLP profile for the 50-150 bp and 150-300 bp intervals). With such small densities, it appears unlikely that homoplasy would strongly affect our data. On the other hand, non-CG primers had a density of selected markers about twice greater for the 50-150 bp class (density = 2.5 loci/10 bp) than for the 150-300 bp class (1.2 loci/10 bp). In addition, a large number of potential markers were not scored for these primers, because they presented peaks either too small or unclear. Such discarded markers were rare for CG-primers. Overall, P5% remained low in albatrosses whatever the range of fragment size considered.
References
Pompanon, F., Bonin, A., Bellemain, E. & Taberlet, P. 2005 Genotyping errors: causes, consequences and solutions. Nature Rev. Genet.6, 847-859.
Vekemans, X., Beauwens, T., Lemaire, M. & Roldan-Ruiz, I. 2002 Data from amplified fragment length polymorphism (AFLP) markers show indication of size homoplasy and of a relationship between degree of homoplasy and fragment size. Mol. Ecol.11, 139-151.
Table 1 Polymorphism in wandering and Amsterdam albatrosses for different ranges of AFLP fragment size.
Size range (base pairs) / number of loci / % polymorphic loci050-100 / 65 / 3.1
100-150 / 68 / 1.5
150-200 / 44 / 9.1
200-250 / 28 / 3.6
250-300 / 17 / 23.5
>300 / 12 / 0