Integrative Biology 200A Spring 2006University of California, Berkeley Kipling Will“PRINCIPLES OF PHYLOGENETICS”2 Feb.- ? Given a chosenalgorithm and a set of coded characters with their assumed homology (alignment forDNA) the set of trees is given. Different tree generating methods often result in the sameor similar relationships. However, changing characters and character coding usuallychanges the actual meaning and interpretation of the evolution of the group (think abouttransformation series, hypothetical ancestors and branch lengths).-Hennig's Auxillary Principle - Assume homology in the absence of contrary evidence.-Conjectural homology assessment (similarity) prior to cladistic analysis andcorroborated homology assessment (homology [process] = synapomorphy [pattern]) andhomoplasy [pattern not due to a common process] after cladistic analysis.-Characters = Transformation series= columns in matrix-Character states= cell entries in matrix-Characters states are the alternative forms of characters observed in the semaphorant.-Character states are sub-grouping of characters, and characters are in turn states at ahigher-level or dependent on states at a higher level.-When the analysis results in corroboration of our initial hypotheses all superordinatecharacters remain unchallenged.-When the analysis results in a pattern of homoplasy, i.e., our initial assumption ofhomology was wrong, the error could be at any level. We may decide that re-coding isappropriate.-At some level no characters are absolutely independent. The level of non-independencevaries along a gradient from negligible to covariance.-Coding is an abstraction of observations.-Coding is a state-ordering process.More on Characters and Character CodingTo Reiterate and expand some ideasWhy are characters as, or more important than the method of analysis-Methods require discrete states.All featuresDescribableHistorical InformationNo Historical InformationA perfect matrix would only beproperly described characters fromthis upper quadrant where homology= synapomorphy.In reality a matrix in made of asample from all describable andvariable characters and so somemislead us and some show a patternof homoplasy.aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa1aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaUseful for identificationVariablePolaritySome methods used to determining a priori polarity:Current Outgroup analysis method- Distinguishing ancestral from derived or the idea of“polarizing” characters originated early in phylogenetics and was central to Hennig's (1966)phylogenetic method and reconstruction methods developed by Wagner (1961, 1969), etc. It wasessential to identify primitive vs. derived character states prior to tree construction. Establishingcharacter state polarity prior to analysis in many papers gave rise to a misconception that it is necessaryto “polarize” characters (e.g. Maddison et al., 1984). Determination of character polarity prior to cladisticanalysis as it is now implemented is usually not desirable.1. “Traditional” Outgroup comparison- Select an OTU or set of OTUs that is/are outside the in-group,but closely related to it, to be the outgroup(s) (best if it includes the sister group). Assume characterstates in outgroup are ancestral.Problems??2. Hypothetical ancestor, sometimes as a “ground-plan” is constructed based on a composite idea ofoutgroup taxa and especially the notion that common equals primitive.Problems??3. Embryological criteria- Application of modified Von Baer's law or as Haeckel proposed "ontogenyrecapitulates phylogeny." General, primitive, ancestral characters appear in embryo before derivedGills---->No GillsProblems??4. Paleontological criterion- Assume older fossils exhibit more ancestral characters.Problems??5. Chorological progression- Species nearer the center of origin of the taxon have the primitive characterstates.Problems??- putative outgroup taxa are included in the analysis and thenetwork is rooted between the ingroup and outgroup(s), and then “character polarity” is based onoptimization of the character on a particular tree topology. This method avoids incorporation ofpreconceived bias into the analysis, allows testing of the monophyly of the ingroup (if more than oneoutgroup is employed). This method was first proposed by Farris (1972).Problems??Direction of character change.2No polarity = 1 0Outgroup 0T1 0T2 0T3 1T4 1(Outgroup (T1( T2, T3))) = (0(0(1,1))) This implies 0 1OutgroupT1T2T3Kinds of Characters and coding exercises:Binary characterMultistate characterNonadditive (=unordered) multistate character adjacency.Additive (=ordered) multistate character- adjacencyBinary and mixed coding of Additive multistate character-Unit coded characters- two state character- 0,1- more than two states- 0,1,2...; ACGT- No set character state Same number ofsteps between any two states.Character with state-to-state specified such that inthe analysis a violation of the ordering cost more steps.012The same logic that is used to establish characters and character states and the hierarchical relationship ofcharacters is the same at this level. You must be explicit about assumptions!Hierarchies and other complexrelationships between character states can be represented in the coding. [a.k.a Probable pathways models,character state trees] - Binary characters that are grouped in sets to define acomplex configuration. Can be used to represent reticulate patterns.Can we justify setting character order (=make them additive or set adjacency)?3Mixed additive codingMore explicit Evolutionary models: (Irreversible charactersDollo charactersstep-matrix, or cost matrixThe triangle inequality(a.k.a. Multistate hierarchic coding or linear nonredundant coding)Typically implemented by the software)- Multiple gains allowed, no losses- Multiple losses allowed, multiple gains not.Implicit in all of these are a character state (Sankoff, 1975), assigning coststo changes.Unordered Ordered Irreversible0123 0123 0 1 2 30|0111 0|0123 0| 0 1 2 31|1011 1|1012 1| 0 1 22|1101 2|2101 2| 0 13|1110 3|3210 3| 0-Step matrices can be used for any number of transformation or weighting schemes, evenasymmetrical ones and ones with non-zero diagonals (cost for not changing), e.g., in the example belowtransversions cost more.ACG
View Full Document
Unlocking...