A COMPARISON OF 2 METHODS FOR CONSTRUCTING EVOLUTIONARY DISTANCES FROM A WEIGHTED CONTRIBUTION OF TRANSITION AND TRANSVERSION DIFFERENCES

Since the initial work of Jukes and Canter (1969), a number of procedu res have been developed to estimate the expected number of nucleotide substitutions corresponding to a given observed level of nucleotide di fferentiation assuming particular evolutionary models. Unlike the prop ortion of different sites, the expected number of substitutions that w ould have occurred grows linearly with time and therefore has had grea t appeal as an evolutionary distance. Recently, however, a number of a uthors have tried to develop improved statistical approaches for gener ating and evaluating evolutionary distances (Schoniger and von Haesele r 1993; Goldstein and Pollock 1994; Tajima and Takezaki 1994). These s tudies clearly show that the estimated number of nucleotide substituti ons is generally not the best estimator for use in reconstruction of p hylogenetic relationships. The reason for this is that there is often a large error associated with the estimation of this number. Therefore , even though its expectation is correct (i.e., on average the expecte d number of substitutions is proportional to time-but see Tajima 1993) : it is not expected to be as useful as estimators designed to have a lower variance.