Beachum, C.*; Strange, R.: an evaluation of phylogeny inference programs and packages

Phylogenetic inferences from large data sets can be problematic, with conflict between precision and minimal computer time determining the design of computer packages. We analyzed two large DNA data sets obtained from GenBank and a smaller homoplasy-rich morphological data set to demonstrate inconsistencies among three phylogeny building packages (PHYLIP, HENNING86, and PAUP*4.0). The sequence data was analyzed under the criterion of maximum parsimony, maximum likelihood, neighbor joining, and bootstrap where applicable. HENNING86 performed well as a parsimony engine, consistently finding the shortest tree in the least amount of time. However it is the oldest of the four packages and is the only one not offered for Unix, Windows, and Mac operating systems. Random Cladistics is a free upgrade that allows many different analyzing methods. PHYLIP3.5 is a free package that can perform many different analyzing methods. It is slow and retains an upward limit of only 100 trees, regardless of the number of equally parsimonious trees. Despite its problems, PHYLIP is free, has the best online help, and is a well integrated package. An alpha version of PHYLIP3.6, released this year, is still slow. It is more plastic and has added an adjustable maximum number of trees. PAUP*4.0 is the most refined of the four programs. It has the greatest number of available options. It is relatively quick and easy to navigate in the point and click format. However, the available options may be daunting to novice users. PAUP* is the best overall package for serious researchers. Researchers and educators need to consider their requirements before committing themselves to specific programs.