Relative efficiencies of the maximum likelihood, maximum parsimony, and neighbor-joining methods for estimating protein phylogeny.

The relative efficiencies of the maximum likelihood (ML), maximum parsimony (MP), and neighbor-joining (NJ) methods for protein phylogeny in obtaining the correct tree topology were studied by using computer simulation. Furthermore, the robustness of the methods against departures from the assumed underlying model was studied. The following conclusions were suggested: (1) While the MP method is sensitive to the violation of constant evolutionary rate among lineages, the ML and NJ methods are robust. (2) The NJ method is efficient even if the evolutionary rate differs among lineages, but it is prerequisite to estimate the distance matrix correctly in this case. (3) The ML method is generally (with some exceptions) robust with respect to violations of the assumed model for amino acid substitution. (4) The ML and MP methods are robust to heterogeneity of evolutionary rate among sites, while the NJ method is sometimes not efficient in the heterogeneous case, unless the heterogeneity is taken into account in estimating the multiple-hit effect.