getParams: Calculate parameters to use as input for HMM

In leekgroup/derfinder: Differential expression analysis of RNA-seq data at base-pair resolution

Description

Assumes that the moderated t statistics obtained by fitting a linear model to each nucleotide come from a Gaussian mixture distribution, where the four distributions in the mixture represent distributions of t statistics from "underexpressed," "overexpressed," "equally expressed," and "not expressed" nucleotides. getParams estimates the parameters of each of the sub-distributions, as well as the percentage of the mixture distribution each contributes, in order to use these parameters to fit a Hidden Markov Model that classifies the nucleotides.

Usage

getParams(tstats, plots = FALSE, plotfile = NULL, verbose = F)

Arguments

tstats	Vector containing all moderated t statistics obtained using getTstats.
plots	if TRUE, create diagnostic plots as parameters are estimated
plotfile	Optional string giving a location and PDF file name to which plots should be written, if plots = TRUE. If NULL, plots are created in the available graphics device.
verbose	If TRUE, periodic messages are printed onscreen during estimation.

Details

The standard pipeline here is to feed the output from getParams directly into getRegions using the "HMM" option.

Value

A list with elements

params	list with elements mean and sd, both 4-item vectors. mean gives the respective means of the "not expressed," "equally expressed," "overexpressed," and "underexpressed" distributions; sd gives their respective standard deviations.
stateprobs	vector of percentages of the mixture distribution that come from the not expressed," "equally expressed," "overexpressed," and "underexpressed" distributions, respectively. It is assumed that "overexpressed" and "underexpressed" t statistics comprise equal percentages of the mixture.

Author(s)

Alyssa Frazee

See Also

getRegions

leekgroup/derfinder documentation built on May 20, 2019, 11:30 p.m.