Large data sets, large sets of trees, and how many brains? Visualization and comparison of phylogenetic hypotheses inferred from rDNA in Chalcidoidea (Hymenoptera)

When large data sets are involved in phylogenies, trivial issues may become central problems. Tree comparison and visualization are two of these issues addressed here. A data set for 18S and 28S D2 through D5 rDNA covering 525 taxa representing all 19 families of Chalcidoidea (Hymenoptera) (65 subfamilies and 267 genera) and five outgroup superfamilies (17 families and 51 genera) was produced and analysed at University of California, Riverside. Sets of trees were produced by MrBayes (1 tree), RaxML (5 trees with different settings for alignment and partitions) and TNT (two outputs ranging from 30,000 to 150,000 trees). We used two hash algorithms, HashRF and HashCS, to detect patterns in this tree set, including topological similarity and partial consensus. The approach described here aims at identifying 1) levels of information about relationships below subfamily level across the obtained trees that cannot be identified on the strict consensus and 2) groups that undermine the resolution of the consensus. We discuss briefly some of the patterns observed.