Rd. Reed et Fah. Sperling, Interaction of process partitions in phylogenetic analysis: An example from the swallowtail butterfly genus Papilio, MOL BIOL EV, 16(2), 1999, pp. 286-297
In this study, we explored how the concept of the process partition may be
applied to phylogenetic analysis. Sequence data were gathered from 23 speci
es and subspecies of the swallowtail butterfly genus Papilio, as well as fr
om two outgroup species from the genera Eurytides and Pachliopta. Sequence
data consisted of 1,010 bp of the nuclear protein-coding gene elongation fa
ctor-1 alpha (EF-1 alpha) as well as the entire sequences (a total of 2,211
bp) of the mitochondrial protein-coding genes cytochrome oxidase I and cyt
ochrome oxidase II (COI and COII). In order to examine the interaction betw
een the nuclear and mitochondrial partitions in a combined analysis. we use
d a method of visualizing branch support as a function of partition weight
ratios. We demonstrated how this method may be used to diagnose error at di
fferent levels of a tree in a combined maximum-parsimony analysis. Further,
we assessed patterns of evolution within and between subsets of the data b
y implementing a multipartition maximum-likelihood model to estimate evolut
ionary parameters fur various putative process partitions. COI third positi
ons have an estimated average substitution rate more than 15 times that of
EF-1 alpha, while COII third positions have an estimated average substituti
on rate more than 22 times that of EF-1 alpha. Ultimately, we found that al
though the mitochondrial and nuclear data were not significantly incongruen
t, homoplasy in the fast-evolving mitochondrial data confounded the resolut
ion of basal relationships in the combined unweighted parsimony analysis de
spite the fact that there was relatively strong support for the relationshi
ps in the nuclear data. We conclude that there may be shortcomings to the m
ethods of "total evidence" and "conditional combination" because they may f
ail to detect or accommodate the type of confounding bias we found in our d
ata.