CONSIDERATIONS FOR MEASURING GENETIC-VARIATION AND POPULATION-STRUCTURE WITH MULTILOCUS FINGERPRINTING

Multilocus DNA fingerprinting provides a cost-effective means to rapid ly assay genetic variation at many loci. While this makes the techniqu e particularly attractive for studies of evolution and conservation bi ology, fingerprint data can be difficult to interpret. Measurement err ors inherent with the technique force investigators to group similar-s ized alleles (bands) into discrete bins before estimating genetic para meters. If too little error is accounted for in this process homologou s alleles will not be grouped in a common bin, whereas overestimated e rror can produce bins With homoplasic alleles. We used simulations and empirical data for two frog species (Rana luteiventris and Hyla regil la) to demonstrate that mean band-sharing ((S) over bar(xy)) and heter ozygosity ((H) over bar(E)) are a function of both bin width and band profile complexity (i.e. number and distribution of bands). These esti mators are also sensitive to the number of lanes included in the analy sis when bin width is wide and a floating bin algorithm is employed. M ultilocus estimates of (H) over bar(E), were highly correlated with (S ) over bar(xy) and thus provide np additional information about geneti c variation. Estimates of population subdivision ((F) over cap and <(P hi)over cap>(ST)) appeared robust to changes in bin size. We also exam ined the issue of statistical independence for band-sharing data when comparisons are made among all samples. This analysis indicated that t he covariance between band-sharing statistics was very small and not s tatistically different from zero. We recommend that sensitivity analys es for bin size be used to improve confidence in the biological interp retation of multilocus fingerprints, and that the covariance structure for band-sharing statistics be examined.