Fig. 6.

Phylogenies estimated from simulated highly divergent data sets (average pairwise sequence identity 13–17%). Top: true tree used to simulate data. Left: sequence–structure model estimates. Right: MAFFT estimates. Central green circle indicates correct topology, whereas red, blue, and yellow identify correct pairs where mismatches are made. The sequence–structure model estimates the correct toplogy in 6 of 10 simulations and preserves correct pairings of the proteins in all but one. MAFFT produces the correct topology in only one data set and in all other cases matches pairs incorrectly.