PhastBias performs a phylo-HMM analysis which assesses the evidence for GC-biased gene conversion (gBGC) on a particular branch of the tree.
phastBias(align, mod, foreground = NULL, do.bgc = TRUE, bgc = 3,
estimate.bgc = FALSE, bgc.expected.length = 1000,
estimate.bgc.expected.length = FALSE, bgc.target.coverage = 0.01,
estimate.bgc.target.coverage = TRUE, sel = -2.01483,
cons.expected.length = 45, cons.target.coverage = 0.3,
estimate.scale = FALSE, post.probs = TRUE)An msa object representing an alignment
An object of type tm representing the neutral
nucleotide substitution model.
A character string giving the name of a branch (or a label given to several branches) indicating which branch should be in the foreground. The foreground branch is where gBGC is tested.
If FALSE, do not model GC-biased gene conversion
Initial value for gBGC parameter B
If FALSE, do not optimize the gBGC parameter,
just hold it at its initial value.
Initial value for expected length of gBGC tract lengths.
If FALSE, do not optimize
the transition rate out of gBGC states (which determines the distribution
of gBGC tract lengths)
Initial value for prior expected target coverage of gBGC tracts (as a fraction between 0 and 1).
If FALSE, constrain the rates
into and out of gBGC state so that bgc.target.coverage does not change.
Set the scaling factor for the conserved state. This is a population genetic parameter which translates to a scaling factor of sel/(1-exp(-sel)). The default value of s=-2.01483 translates to a scaling factor of 0.31 in the background branches.
Set the expected length of conserved elements.
Set the target coverage for conserved elements.
If TRUE, estimate a scaling factor for the
branch lengths in all states.
If TRUE, return value will include a data frame
containing posterior probabilities for every position in the alignment
and every state. Set to FALSE to suppress.
A list containing parameter estimates, a features object predicting which part of the alignments have gBGC probability > 0.5, and a data frame with posterior probabilities at all positions (if post.probs==TRUE)
PhastBias utilizes a HMM with the following states: neutral, conserved, neutral with gBGC, and conserved with gBGC. The scaling factor between conserved/neutral, the strength of gBGC, and the transition rates between states can be configured. It produces posterior probabilities for each state for every column of the alignment, or a set of gBGC "tracts" giving the regions where gBGC is predicted (by thresholding the posterior probability at 0.5).