Lecture 9 - Phylogenies (continued)

included on exam 2

pruning

the “clipping” of a tree

the removal of taxa

the three types of tree building

parsimony

maximum likelihood

Bayesian

parsimony

the principle that tells us to choose the simplest scientific explanation that fits the evidence

choose the simplest solution, with the fewest changes

rarely ever actually used

e.g. if three species are compared by genetic sequence, the two species with the fewest nucleotide base differences will be sister taxa, and the one with more nucleotide differenes will be the outgroup

maximum likelihood

methods used to estimate phylogenetic trees

the probability of certain data if given a tree (all trees taken into account)

essentially, parsimony with numbers to back it up, e.g. data resampling

contains node support value numbers from 0-100

Bayesian method

a method of tree building that involves posterior probability

contains node support value numbers from 0-1

nucleotide substitution models

models used to calculate the likelihood of phylogenetic trees using multiple sequence alignment data

JC, F81, HKY, maximum likelihood

JC model

a nucleotide substitution model

assumes that A, T, C, and G all occur at ~25% frequency, and substitution rates are equal for all base pairs

don’t really worry about it

F81 model

a nucleotide substitution model

assumes variable base pair frequency (not all 25%), and substitution rates are equal for all base pairs

don’t really worry about it

HKY model

a nucleotide substitution model

assumes variable base pair frequency, and substitution rates are not equal between base pairs

purine→purine or pyrimidine→pyrimidine substitutions are easier, occur more frequently

purine→pyrimidine or pyrimidine→purine substitutions are harder, occur less frequently

data resampling methods

bootstrapping and jacknifing

bootstrapping

a method of data resampling

data is sampled with replacement of the data (replacement meaning put back, not exchanged)

i.e. the same data point can be sampled multiple times

jackknifing

a method of data resampling

the matrix has to get smaller, data is removed from the pile after sampling