Dynamically heterogenous partitions and phylogenetic inference: An evaluation of analytical strategies with cytochrome b and ND6 gene sequences in cranes

Debates over whether molecular sequence data should be partitioned for phyl ogenetic analysis often confound two types of heterogeneity among partition s. We distinguish historical heterogeneity (i.e., different partitions have different evolutionary relationships) from dynamic heterogeneity (i.e., di fferent partitions show different patterns of sequence evolution) and explo re the impact of the latter on phylogenetic accuracy and precision with a t wo-gene, mitochondrial data set for cranes. The well-established phylogeny of cranes allows us to contrast tree-based estimates of relevant parameter values with estimates based on pairwise comparisons and to ascertain the ef fects of incorporating different amounts of process information into phylog enetic estimates. We show that codon positions in the cytochrome b and NADH dehydrogenase subunit 6 genes are dynamically heterogenous under both Pois son and invariable-sites + gamma-rates Versions of the F84 model and that h eterogeneity includes variation in base composition and transition bias as well as substitution rate. Estimates of transition bias and relative-rate p arameters from pairwise sequence comparisons were comparable to those obtai ned as tree-based maximum likelihood estimates. Neither rate-category nor m ixed-model partitioning strategies resulted in a loss of phylogenetic preci sion relative to unpartitioned analyses. We suggest that weighted-average d istances provide a computationally feasible alternative to direct maximum l ikelihood estimates of phylogeny for mixed-model analyses of large, dynamic ally heterogenous data sets. (C) 1999 Academic Press.