FUBAR : Faste Unbiased Bayesian AppRoximation

What question does this method answer?

Which site(s) in a gene are subject to pervasive, i.e. consistently across the entire phylogeny, diversifying selection?

Recommended Applications

The phenomenon of pervasive selection is generally most prevalent in pathogen evolution and any biological system influenced by evolutionary arms race dynamics (or balancing selection), including adaptive immune escape by viruses. As such, FUBAR is ideally suited to identify sites under positive selection which represent candidate sites subject to strong selective pressures across the entire phylogeny.

FUBAR is our recommended method for detecting pervasive selection at individual sites on large (> 500 sequences) datasets for which other methods have prohibitive runtimes, unless you have access to a computer cluster.

Brief description

Perform a Fast Unbiased AppRoximate Bayesian (FUBAR) analysis of a coding sequence alignment to determine whether some sites have been subject to pervasive purifying or diversifying selection. There are three methods for estimating the posterior distribution of grid weights: collapsed Gibbs MCMC (faster), 0-th order Variation Bayes approximation (fastest), full Metropolis-Hastings (slowest).

Input

1. A FASTA sequence alignment.
2. A phylogenetic tree in the Newick format

Note: the names of sequences in the alignment must match the names of the sequences in the tree.

Output

A JSON file with analysis results (http://hyphy.org/resources/json-fields.pdf).

A custom visualization module for viewing these results is available (see http://vision.hyphy.org/FUBAR for an example)

Further reading

http://hyphy.org/methods/selection-methods/#FUBAR

Tool options

--code             Which genetic code to use

--grid             The number of grid points
                    Smaller : faster
                    Larger : more precise posterior estimation but slower
                    default value: 20

--method           Inference method to use
                        Variational-Bayes : 0-th order Variational Bayes approximation; fastest [default]
                        Metropolis-Hastings : Full Metropolis-Hastings MCMC algorithm; orignal method [slowest]
                        Collapsed-Gibbs  : Collapsed Gibbs sampler [intermediate speed]


--chains           How many MCMC chains to run (does not apply to Variational-Bayes)
                        default value: 5

--chain-length     MCMC chain length (does not apply to Variational-Bayes)
                        default value: 2,000,000

--burn-in          MCMC chain burn in (does not apply to Variational-Bayes)
                        default value: 1,000,000

--samples          MCMC samples to draw (does not apply to Variational-Bayes)
                        default value: 1,000

--concentration_parameter
                    The concentration parameter of the Dirichlet prior
                    default value: 0.5