Shared single copy genes are generally reliable for inferring phylogenetic relationships among polyploid taxa

Polyploidy, or whole-genome duplication, is expected to confound the
inference of species trees with phylogenetic methods for two reasons.
First, the presence of retained duplicated genes requires the
reconciliation of the inferred gene trees to a proposed species tree.
Second, even if the analyses are restricted to shared single copy genes,
the occurrence of reciprocal gene loss, where the surviving genes in
different species are paralogs from the polyploidy rather than orthologs,
will mean that such genes will not have evolved under the corresponding
species tree and may not have gene trees that allow inference of the
species tree. Here we analyze three different ancient polyploidy events,
using synteny-based inferences of orthology and paralogy to infer gene
trees from more than 17,000 sets of homologous genes. We find that the
simple use of single copy genes from polyploid organisms provides
reasonably robust phylogenetic signals, despite the presence of reciprocal
gene losses. Such gene trees are also most often in accord with the
inferred species relationships inferred from maximum likelihood models of
gene loss after polyploidy: a completely distinct phylogenetic signal
present in these genomes. As seen in other studies, however, we find that
methods for inferring phylogenetic confidence yield high support values
even in cases where the underlying data suggest meaningful conflict in the
phylogenetic signals.