Models and properties of character information for phylogenetic analysis, with examples from molecular data and slipper orchids (Cypripedioideae: Orchidaceae)

Information use and evaluation in phylogenetics must be consistent with logical goals (for example, utility and conservation, not accuracy). The information space for a given phylogenetic question is defined by uncertain boundaries of parameters such as the patterns of taxon, character, and character-state sampling; as such, systematically useful information is sampled nonrandomly without an underlying assumption of an infinite distribution of such data. Parsimony analysis explicitly summarizes evidence into a hierarchical form that is both utilitarian and information-conservative. Accuracy in terms of closeness to a true tree cannot be directly addressed by any method that purports to hypothesize phylogenetic relationships. Nevertheless, congruence analyses between information sub-spaces is a justifiable indirect approach to examining the commonality of evidence present in each, which may in turn reflect upon approximation of true trees. Both component-oriented and character-oriented congruence techniques are informative regarding the strength of data interaction; the former focuses on topological consensus of independently-derived phylogenetic hypotheses, whereas the latter examines character concordance in a single phylogenetic reconstruction derived from all available evidence. Such understanding and study of phylogenetically useful information becomes particularly important when molecular data are involved. DNA, for example, presents a character system that is limited in terms of available states. Phylogenetic reconstruction of characters that display only four possible states over an entire gene sequence may limit the resolving power of a most-parsimonious summmarization. Evolutionary models of character-state change for nucleotide sequences and nucleotide strings (restriction endonuclease cleavage sites) have been used to erect weights that rationally reflect transformational asymmetries observed in these data. However, these weighting schemes are so near to an assumption of equal likelihood of character-state change that the latter may be a more robust (and certainly less ad hoc) criterion for phylogenetic analysis. Similar character-state weighting issues arise with amino acid sequence data; the twenty amino acids incorporated into proteins may not transform from state to state with equal likelihood given constraints implied by the genetic code. All of these issues are explored using plant systems, with a summary example of congruence between molecular and morphological hypotheses of slipper orchid relationships (Cypripedioideae: Orchidaceae).