Lecture 13 - Maximum Likelihood Flashcards
what does maximum likelihood maximize in the context of phylogenetic trees
P(data | tree, model)
- the probability of observing the data given the model (tree and model of evolution)
what does L(model | data) mean
given a model of sequence evolution, find the tree and parameters which maximizes the likelihood of the observed data (alignment)
define: probability
predicted result based on the assumed model
define: likelihood
predict the model based on the observed set of results
what is the data for calculating maximum likelihood
MSA
what is the model of evolution used for calculating maximum likelihood
one of JC, K2P, GTR, etc.
how do you find the total likelihood that one sequence becomes another
take the product of the likelihoods of each column in the alignment
how do you incorporate trees as part of the ML model
if you have more than 3 sequences, each term in the likelihood calculation is expanded to incorporate tree topologies and branch lengths; the probability of base pair changes on a very short branch length is low
does maximum likelihood generate trees
no, it only compares them
what is the likelihood of a tree
the product of the site likelihoods
what are the strengths of maximum likelihood
- it gives the correct tree given the model is correct
- it has accurate branch lengths
- it uses all data
what are the weakness of maximum likelihood
- it is computationally intensive
- it is inconsistent if the wrong model is chosen
what is the model of evolution in parsimony
the number of observed changes
what is long branch attraction
the phenomenon in which two taxa are placed together on a tree, not because they are closely related, but because they both have many mutations
what does maximum likelihood account for in terms of rates of evolution under the correct model
it accounts for varying rates of evolution across different branches
how do you pick the most appropriate model for maximum likelihood
as before, AIC or similar measures can pick the ‘best’ model to use; choose the model with the lowest AIC
what does Bayesian maximize
the chance of seeing the tree given the data
what is the Bayesian framework for likelihood models
P(Tree | Data)
what are pipelines
combination of multiple software/tools into a single workflow
