January 11-15, 2003
Town & Country Convention Center
San Diego, CA
Bioinformatics: Software
Computer: Poster and Demo
In the previous century methods that came to be called phenetic clustering were advocated for biological classification. These were best able, it was argued, to represent descriptive information on the characteristics of organisms and the degrees of difference between taxa. But biological classification is now phylogenetic; phenetic methods have disappeared from systematics. Naturally the principal reason for this is the central role that evolution plays in biology, but it is not the only reason. Phylogenetic methods of classification, as it turned out, are better able to represent information on characteristics and degrees of difference than can phenetic methods. This is a result of the logical properties of the methods and would hold true even if there had been no phylogeny (Farris, 1979). Phylogenetic hierarchies are fully diagnosable by the synapomorphies (shared derived state changes) that argue for them. In addition, data can suggest more than one equally most-parsimonious hierarchy, and resampling methods have been developed to assess support for particular reconstructions. Researchers outside of systematics have apparently remained unaware of these developments, however. The clustering techniques now used for analysis of microarray expression profile data are essentially those once advocated for phenetic classification (e.g., http://ep.ebi.ac.uk/EP/). In this presentation we will explore the utility of phylogenetic methods as tools for analyzing expression profile data. Reference: Farris, J. S. (1979). The information content of the phylogenetic system. Systematic Zoology, 28: 483-519.