EXPERIMENTAL-ANALYSIS OF VARIANCE FOR DNA HYBRIDIZATION .2. PRECISION

Using DNAs from the Virginia opossum (Didelphis virginiana), we estimated the variance components for two classes of replicate hybrids: different drivers matched to the same tracer and different homoduplexes made from tracers matched to identical drivers. A nested analysis of variance (ANOVA) was used to partition total variance among four levels: Individuals, extracts, preparations, and different aliquots from the same preparation. The variance contributed by these levels depended on the kind of hybrid replicate (driver or tracer) and on the index of thermal stability (T(mode), T(m), T50H, or Normalized Percentage Hybridization). For replicate drivers, significant variance contributions were made by (1) individuals to T(m), (2) extracts to T(mode) and NPH but not T(m), and (3) different preparations to NPH. The composite T50H measure calculated from both T(m) and NPH revealed effects from both constituent indices. For replicate tracers, preparation error was the single most consistent effect across all indices, followed by extract effects for those indices that incorporated a measure of percent hybridization (T50H, NPH). Total variance of the four indices was qualitatively similar for both drivers and tracers: T(mode) ranked lowest, followed in order by T(m), T50H, and NPH, with the variance of NPH being as much as 100 times greater than for T(mode). These results provide guidelines for the design of experiments to generate DNA hybridization-based phylogenies and to assess their robustness with bootstrapping. Replicate drivers for a distance matrix based on T(m) should use different individuals, whereas one based on T(mode) could minimally use different extracts from the same individual. Thus, T(mode) may be the index of choice for DNA hybridization experiments when material, time, and money are limited.