Repitools (version 1.18.0)

methylEst: Function to derive regional methylation estimates.

Description

Posterior mean and variance for the regional methylation level are derived for all genomic regions. Credible intervals can be computed either numerically from the posterior marginal distribution or by computing them on logit scale and transferring them back.

Usage

methylEst(x, verbose=FALSE, controlCI = list(compute = FALSE, method = "Wald",  level = 0.95, nmarg = 512, ncpu = NULL))

Arguments

x

Object of class BayMethList.

verbose

Boolean indicating whether the methylEst function should run in a verbose mode (default 'FALSE').

controlCI

list defining whether credible intervals should be derived.

-

compute logical. If 'TRUE' credible intervals are derived. (default FALSE)

-

method There are three possible types of credible intervals that can be chosen if a uniform prior, e.g. Beta(1,1), is chosen: 'Wald' (default), 'HPD', 'quantile'. The Wald-type intervals are the fastest to compute. The are calculated on logit scale and then transferred back. Due to numerical integration of the posterior marginal posterior distributions, the computation of highest posterior density (HPD) interval and quantile-based interval is computationally more expensive. However, in our applications HPD intervals provided best coverage.

Note, using a beta mixture or a Dirac-beta-Dirac (DBD) mixture as prior distribution for the methylation level only method="quantile" is available.

-

level numerical value defining the credible level. Default: 0.95.

-

nmarg Number of points at which the posterior marginal is evaluated (only relevant for method="quantile" or method="HPD").

-

ncpu Number of CPUs on your machine you would like to use in parallel. If ncpu is set to NULL, half of the CPUs will be used on machines with a maximum of four CPUs, and 2/3 will be used if more are available.

Value

A BayMethList object where the slot methEst is filled with a list containing the following elements:

mean

Matrix where the number of columns equals the number of samples of interest. Each column contains the posterior mean methylation level for each bin.

var

Matrix where the number of columns equals the number of samples of interest. Each column contains posterior variance for each bin.

ci

List with length equal to the number of samples of interest. Each list element contains a matrix where the first column contains the lower CI bound and the second column the upper CI bound.

W

Matrix where the number of columns equals the number of samples of interest. Each column contains the normalisation factor of the posterior marginal distribution for each bin.

al

Matrix where the number of columns equals the number of samples of interest. Each column contains the prior shape parameter for each bin

bl

Matrix where the number of columns equals the number of samples of interest. Each column contains the prior scale parameter for each bin

Details

The posterior mean and the variance are analytically available and therefore straightforward to efficiently compute; Wald-based credible intervals are obtained on logit scale and then back-transferred to ensure values withing 0 and 1. HPD and quantile-based credible intervals are computed by numerical integration of the posterior marginal distribution.

Examples

    if(require(BSgenome.Hsapiens.UCSC.hg18)){
        windows <- genomeBlocks(Hsapiens, chrs="chr21", width=100, spacing=100)
        cpgdens <- cpgDensityCalc(windows, organism=Hsapiens, 
            w.function="linear", window=700)  
        co <- matrix(rnbinom(length(windows), mu=10, size=2), ncol=1)
        sI <- matrix(rnbinom(2*length(windows), mu=5, size=2), ncol=2)
        bm <- BayMethList(windows=windows, control=co, 
            sampleInterest=sI, cpgDens=cpgdens)

        bm <- determineOffset(bm)
        # should take about 3 minutes for both samples of interests with 2 CPUs.
        bm <- empBayes(bm) 
        bm <- methylEst(bm, controlCI = list(compute = FALSE, method = "Wald", 
            level = 0.95, nmarg = 512, ncpu = NULL))
    }