CpGFilter (version 1.1)

CpGFilterICC: CpG filtering method based on intra-class correlation coefficients.

Description

Filter CpGs based on Intra-class Correlation Coefficients (ICCs). ICCs are calculated by fitting linear mixed effects models to all samples including the un-replicated samples. Including the large number of un-replicated samples improves ICC estimates dramatically. The method accommodates any replicate design.

Usage

CpGFilterICC(dat, rep.design, REML = FALSE, logit.transform = TRUE, verbose = TRUE)

Arguments

dat

a matrix of CpG beta-values, row - CpG, column - sample

rep.design

a vector indicating the replicate design, it could be factor, character or numeric vectors. Example - c(1, 2, 3, 4, 4, 4, 5, 5) OR c('S1', 'S2', 'S2', 'S2', 'S1')

REML

If TRUE, Restricted Maximum Likelihood (REML) method will be used; Otherwise, Maximum Likelihood (ML) method will be used. Default is FALSE.

logit.transform

If TRUE, beta-value will be converted into M-value; Default is TRUE.

verbose

If TRUE, print run information

Value

ICCs for all probes

References

Chen J, Just A, et al. CpGFilter:Model-based CpG probe filtering with replicates for epigenome-wide association studies (2016). Bioinformatics, 32(3): 469<U+2013>471

Examples

# NOT RUN {
require(CpGFilter)
# 10 samples replicated twice, 5 samples replicated four times.
rep.design <- c(1:100, 101:110, 101:110, 111:115, 111:115, 111:115, 111:115)
rho <- CpGFilterICC(matrix(rnorm(140*1000), 1000, 140), rep.design, logit=FALSE)
# }