BiSeq (version 1.12.0)

betaRegression: A function to estimate and test a group factor within a beta regression

Description

This function models the methylation level within a beta regression. The first independent variable in formula is tested to be unequal to zero.

Usage

betaRegression(formula, link, object, mc.cores, ...)

Arguments

formula

Symbolic description of the model. For the first independent variable the P value (Wald test) and the effect on methylation is returned. For details see below.

link

A character specifying the link function in the mean model (mu). Currently, "logit", "probit", "cloglog", "log", "loglog" are supported.

object

A BSrel object.

mc.cores

Passed to mclapply.

...

Other parameters passed to the betareg function.

Value

A data.frame containing the position, chromosome, P value, estimated methylation level in group 1 and group 2 and methylation difference of group 1 and group 2.

Details

See betareg function for details.

mclapply

References

Hebestreit, K., Dugas, M., and Klein HU. Detection of significantly differentially methylated regions in targeted bisulfite sequencing Data. In preparation. Bioinformatics. 2013 Jul 1;29(13):1647-53. See also reference list in the documentation of betareg.

See Also

betareg

Examples

# load RRBS data, subset to save time, find CpG clusters and smooth methylation data:
data(rrbs)
rrbs.small <- rrbs[1:1000,]
rrbs.clust.unlim <- clusterSites(object = rrbs.small,
                                 groups = colData(rrbs)$group,
                                 perc.samples = 4/5,
                                 min.sites = 20,
                                 max.dist = 100)

ind.cov <- totalReads(rrbs.clust.unlim) > 0
quant <- quantile(totalReads(rrbs.clust.unlim)[ind.cov], 0.9)
rrbs.clust.lim <- limitCov(rrbs.clust.unlim, maxCov = quant)

# with a small subset to save calculation time:
rrbs.part <- rrbs.clust.lim[1:100,]

predictedMeth <- predictMeth(object=rrbs.part)

betaResults <- betaRegression(formula = ~group, link = "probit",
                              object = predictedMeth, type="BR")