Dubious resolution and support from published sparse supermatrices: the importance of thorough tree searches

SIMMONS, M.; GOLOBOFF, P.

MOLECULAR PHYLOGENETICS AND EVOLUTION

ACADEMIC PRESS INC ELSEVIER SCIENCE

Lugar: Amsterdam; Año: 2014 vol. 78 p. 334 - 348

1055-7903

We re-analyzed 10 sparse supermatrices wherein the original authors relied primarily or entirely uponmaximum likelihood phylogenetic analyses implemented in RAxML and quantified branch support usingthe bootstrap. We compared the RAxML-based topologies and bootstrap values with both superficial- andrelatively thorough-tree-search parsimony topologies and bootstrap values. We tested for clades thatwere resolved by RAxML but properly unsupported by checking if the SH-like aLRT equals zero and/orif the parsimony-optimized minimum branch length equals zero. Four of our conclusions are as follows.(1) Despite sampling nearly 50,000 characters, highly supported branches in a RAxML tree may beentirely unsupported because of missing data. (2) One should not rely entirely upon RAxML SH-like aLRT,RAxML bootstrap, or superficial parsimony bootstrap methods to rigorously quantify branch support forsparse supermatrices. (3) A fundamental factor that favors thorough parsimony analyses of sparsesupermatrices is being able to distinguish between clades that are unequivocally supported by the datafrom those that are not; superficial likelihood analyses that quantify branch support using the bootstrapcannot be relied upon to always make this distinction. (4) The SH-like aLRT and parsimony-optimizedminimum-branch-length tests generally identify the same properly unsupported clades; the latter is amore severe test.