Trees - ML and Bayes Flashcards
What does maximum likelihood do
it tries to infer a phylogenetic tree by finding the tree that maximizes the probability of observing the data, given certain assumptions (models)
How does the maximum likelihood process work
it requires explicitly models of rates of evolutionary change in the relevant characters and states, it begins with one tree (can be hypothesis or model) then goes on to compare the probability of the data for many different trees/models, searching across tree space for the tree that maximizes the likelihood of the data
The maximum likelihood method model includes what 4 things
- tree topology
- branch lengths
- nucleotide frequencies
- a substitution rate matrix/instantaneous rate matrix
What are the 2 models for substitution rate matrix’s
- Jukes-Cantor (JC) model
- Kimura two parameter (K2P) model
Describe the Jukes Cantor model used for maximum likelihood and finding the substitution rate matrix
- it only has one parameter, the substitution rate, and assumes that all substitutions (changes of character state) are equally likely
- the substitution rate is expressed as α (alpha) per time unit.
- given enough time the probability of change to each of the other 3 states and the probability of no change all approach 0.25
- each column or row adds up to zero
Describe the Kimura two-parameter model used for maximum likelihood and finding the substitution rate matrix
this has one rate for transitions (α) and one rate for transversions (β)
- each column or row adds up to zero
What is a faster process: maximum likelihood or maximum parsimony
- maximum parsimony - heuristic methods(traditional searching) is almost always used)
Describe bayesian inferene
- built on likelihood methods
- based on matrix of character states and requires an evaluation of all possible trees or a heuristic sample of them
- it involves beginning with prior probabilities and then sequentially estimating revised posterior probabilities based on the observations actually made
- it calculates the probability of the hypothesis given the data (reverse of ML)
What kind of method does the bayesian inference search involve
markov-chain monte carlo method (MCMC) - this is a random walk simulation–> this leads to convergence on a set of trees for which the likelihoods are similar
Difference between Bayesian inference and maximum likelihood
- ML = finds the tree that maximizes the probability of observing the data given certain assumptions/models
- Bayesian = calculates the probability of the hypothesis given the data
What is the Bayesian inference model sensitive to
priors and model misspecification
What is the order for the methods of tree construction
- cluster analysis
- maximum parsimony
- likelihood methods (ML, BI)