Integrative Biology 200A "PRINCIPLES OF PHYLOGENETICS" Spring 2006 University of California, Berkeley B.D. Mishler April 20, 2006. Phylogenetic Trees IX: Summary -- what are they, really, and what can go wrong? A. Lead-off discussion We have now reached the point in the course where we have covered both the character analysis phase and the cladistic analysis phase of phylogenetics in great detail. So let's step back and look at these phases together in a more integrated, sophisticated way. Here are some important initial questions for discussion: What are phylogenetic trees, really? What do you see when you look closely at a branch? -- the fractal nature of phylogeny (is there a smallest level?) What is the relationship between characters and trees? Characters and OTUs? Characters and levels? -- exemplar coding versus composite coding What can go wrong in the fit between characters and trees? -- random versus caused homoplasy -- epistemological problems versus ontological problems B. Summary of character analysis The process of phylogenetic analysis inherently consists of two phases. First a data matrix is assembled, then a phylogenetic tree is inferred from that matrix. There is obviously some feedback between these two phases, yet they remain logically distinct parts of the overall process. Paradoxically, despite the logical preeminence of data matrix construction in phylogenetic analysis, by far the largest effort in phylogenetic theory has been directed at the second phase of analysis, the question of how to turn a data matrix into a tree. Extensive series of publications have been elaborated to attempt to justify such tree-building approaches as neighbor-joining, maximum likelihood, and Bayesian inference, while ignoring entirely the nature of the data matrix that must underlie any analysis. The reasons for this asymmetry in research on phylogenetic theory are not entirely clear, but it probably has to do with the fact that the problem of tree building may appear simpler, more clear cut. Perhaps it is just a matter of research fashions. For whatever reason, relatively little attention has been paid to the assembly of the data matrix. At stake are each of the logical elements of the data matrix: the rows (what are OTUs?), the columns (what are characters?), and the individual entries (what are character states?). The tree of life is inherently fractal, which complicates the search for answers to these questions. Look closely at one lineage of a phylogeny and it dissolves into many separate lineages, and so on down to a very fine scale. Thus the nature of both OTU's ("operational taxonomic units," the "twigs" of the tree in any particular analysis) and characters (hypotheses of homology, markers that serve as evidence for the past existence of a lineage) change as one goesup and down this fractal scale. Furthermore, there is a tight interrelationship between OTUs and character states, since they are reciprocally recognized during the chacter analysis process. C. Summary of tree building One could easily argue that the first phase of phylogenetic analysis is the most important; the tree is basically just a re-representation of the data matrix with no value added. This is especially true from a parsimony viewpoint, the point of which is to maintain an isomorphism between a data matrix and a cladogram. Under this viewpoint, we should be very cautious of any attempt to add something beyond the data in translating a matrix into a tree! If care is taken to construct an appropriate data matrix to address a particular question of relationships at a given level, then simple parsimony analysis is all that is needed to transform a matrix into a tree. Debates over more complicated models for tree-building can then be seen for what they are: attempts to compensate for marginal data. But what if we need to push the envelope and use data that are questionably suited for a particular problem? More complicated model-based methods (weighted parsimony, ML, and Bayesian inference) can be used to push the utility of data, but need to be done carefully. Both the model itself and the values for the parameters in the model need to be based on solid a priori evidence, not inferred ad hoc solely from the data to be used. D. "Deep" versus "shallow" phylogenetic inference: molecules and morphology The problems faced at different temporal scales are quite distinct (Mishler, 2000. Taxon 49: 661-683). In "shallow" reconstruction problems, the branching events at issue happened a relatively short time ago and the set of lineages resulting from these branching events is relatively complete (extinction has not had time to be a major effect). In these situations the relative lengths of internal and external branches are similar, giving less opportunity for long branch attraction. However, the investigator working at this level has to deal with the potential confounding effects of reticulation and lineage sorting. Characters, at least at the morphological level, may be quite subtle, and at the nucleotide level it is necessary to look very carefully to find rapidly evolving genes (however, such genes are likely to be relatively neutral, thus less subject to adaptive constraints which can lead to non independence). In "deep" reconstruction problems, the branching events at issue happened a relatively long time ago and the set of lineages resulting from these branching events is relatively incomplete (extinction has had a major effect). In these situations, the relative lengths of internal and external branches are often quite different, thus there is more opportunity for long branch attraction, even though there is little to no problem with reticulation and lineage sorting since most of the remaining branches are so old and widely separated in time. Due to all the time available on many branches, many potential morphological characters should be available, yet they may have changed so much as to make homology assessments difficult; the same is true at the nucleotide level, where multiple mutations in the same region may make alignment difficult. Thus very slowly evolving genes must be found, but that very conservatism is caused by strong selective constraints which increases the danger of convergence.a "shallow" reconstruction problema "deep" reconstruction problemtime How will we ultimately connect up
View Full Document
Unlocking...