RECONSTRUCTION OF ORGANISMAL AND GENE PHYLOGENIES FROM DATA ON MULTIGENE FAMILIES - CONCERTED EVOLUTION, HOMOPLASTY, AND CONFIDENCE

The reliability of phylogenies reconstructed from data on multigene families is investigated via simulation. The evolutionary scenario used is a character-based model of a two-gene family in four species in which clocklike divergence is postulated but neither convergence nor reversal is allowed except as a result of recombination and gene conversion. Thus, any homoplasy emerging from parsimony reconstructions from the simulated data matrices can be attributed to concerted evolution. The probabilities of correctly reconstructing two standard trees are estimated by replicate runs of the simulation. One standard tree (the OP or "orthology/paralogy" tree) reflects the true gene genealogy in the absence of concerted evolution; the other (the CE or "concerted evolution" tree) depicts gene relationships under complete homogenization of the gene family. The probability of correct reconstruction of the OP tree declines quickly as concerted evolution increases, but above an intermediate level of concerted evolution the probability of correctly inferring the CE tree increases rapidly. Trees similar but not identical to the correct trees can be reconstructed above or below the critical intermediate level of concerted evolution. Levels of homoplasy and numbers of equally parsimonious minimal trees are maximized, and bootstrap confidence levels are minimized, near this intermediate level of concerted evolution. When reconstructing the correct gene tree is the goal, both consistency indices and bootstrap levels will show misleadingly high values when concerted evolution is high. However, because the correct species tree can be inferred from either the OP or CE tree (in the absence of homoplasy from sources other than concerted evolution), these same measures correlate well with fidelity of reconstructing the species tree.