Lecture 21 Phylogenetics Methods And Likelihood Flashcards
General procedures for Phylogenies
- Choose species
- Collect character data (create data matrix)
- Find optimal tree (e.g., Parsimony or likelihood search
- Root tree (establish character polarities or outgroup rooting)
- Estimate confidence or strength of support (e.g., bootstrapping)
Character polarity determines the
Direction of changes in character states
Computer programs produce
Unrooted trees
Outgroup method
Effectively same as “pulling: down the root on the branch between in-group and outgroup
How do we test the hypotheses of a phylogeny?
Collect more data
Bootstrap
Two ways to look at a tree
- How confident should we be of the entire tree
- How confident should we be of each one (hypothesis) in the tree?
Phylogenetic history must be tested
indirectly
Support for nodes (4)
Consensus trees
Congruence test
Character support
Bootstrapping
Consensus trees
Summarize agreement among different Phylogenies
(Strict consensus and majority rule consensus)
Strict consensus
Conflicts are collapsed on the tree to represent that we do not know 100% what is going on
Majority-rule consensus
Put percentages on the tree to represent how sure we are of its placement
Congruence test
A phylogenetic hypothesis makes predictions about the distribution of other characters
So to test you compare trees from different data sets to see if they are the same or not
Character support
Numbers of synapomorphies
Bootstrapping
Resamples a data set repeatedly to estimate confidence (robustness or repeatability)
Likelihood
A general statistical approach applied to many areas of science
Optimality, way of choosing amongst alternatives
Likelihood equation
L=P(D|M)
Hypothesis =
History (tree topology and branch lengths) plus model of evolution
Model =
Transition rates for character state transformation and for classes of characters (quantifying how the DNA sequence evolves; a general description of how things work)
L = P(D|M, T)
If you hold D and M constant, one can compare the likelihood of different trees
Parsimony is a
Special case of likelihood, where every character has its own model.
The phylogeny that makes the data most likely is preferred.
Branch lengths are important when using
Likelihood to choose a tree (changes are more likely on long branches) but irrelevant to parsimony
How to calculate Likelihood
- Map characters (parsimony)
- Ignore outgroup Branch
- Assign Branch probabilities
- Add scenarios on long branches
- Multiply probabilities
L = P(D|M, T) expanded
L=P(data|model, tree)
