EXPERIMENTAL-ANALYSIS OF VARIANCE FOR DNA HYBRIDIZATION .1. ACCURACY

We used tissues of the Virginia opossum (Didelphis virginiana) to examine the experimental accuracy of DNA hybridization statistics of thermal stability (T(mode), T(m), T50H, and NPH) with respect to systematic biases in counting radioactivity in elution fractions, and column position and loading order of hybrids in the thermal elution device. We failed to detect any change in the mean melting temperatures among five replicate I-125-labeled hybrids counted over 72 h. Furthermore, column position in the automated thermal elution device (TED) did not bias the statistics of aliquots loaded over a few minutes from a single large ''mother'' hybrid. On the other hand, the normalized percentage hybridization (NPH) increased as much as 3-5% for aliquots loaded during 1 h from a similar ''mother'' hybrid. A parallel but less consistent increase was noted for T50H, which incorporates a measure of NPH. This ''NPH effect'' disappeared when hybrids were prepared individually and diluted and loaded in turn-the usual procedure in our laboratory. Replicate distances measured as NPH appear to be sensitive to departures from the normal-distribution assumption of least-squares regression. We recommend that replicate cell values of NPH be transformed to improve their fit to a normal distribution prior to analysis by least-squares phylogenetic algorithms such as those available in Felsenstein's PHYLIP package. Thus, potential sources of inaccuracy in DNA hybridization data can be avoided with simple precautions.