GetResiduals: Get Linear Model Residuals

In TransBioInfoLab/coMethDMR: Accurate identification of co-methylated and differentially methylated regions in epigenome-wide association studies

Get Linear Model Residuals

Description

Remove covariate effects from methylayion values by fitting probe-specific linear models

Usage

GetResiduals(
  dnam,
  betaToM = TRUE,
  epsilon = 1e-08,
  pheno_df,
  covariates_char,
  nCores_int = 1L,
  ...
)

Arguments

dnam	data frame or matrix of methylation values with row names = CpG IDs and column names = sample IDs. This is often the genome-wide array data.
betaToM	indicates if methylation beta values (ranging from [0, 1]) should be converted to M values (ranging from (-Inf, Inf)). Note that if beta values are the input to dnam, then betaToM should be set to TRUE, otherwise FALSE.
epsilon	When transforming beta values to M values, what should be done to values exactly equal to 0 or 1? The M value transformation would yield -Inf or Inf which causes issues in the statistical model. We thus replace all values exactly equal to 0 with 0 + epsilon, and we replace all values exactly equal to 1 with 1 - epsilon. Defaults to epsilon = 1e-08.
pheno_df	a data frame with phenotype and covariates, with variable Sample indicating sample IDs.
covariates_char	character vector for names of the covariate variables
nCores_int	Number of computing cores to be used when executing code in parallel. Defaults to 1 (serial computing).
...	Dots for additional arguments passed to the cluster constructor. See CreateParallelWorkers for more information.

Details

This function fits an ordinary linear model predicting methylation values for each probe from the specified covariates. This process will be useful in scenarios where methylation values in a region or at an individual probe are known a priori to have differential methylation independent of the disease or condition of interest.

Value

output a matrix of residual values in the same dimension as dnam

Examples

   data(betasChr22_df)

   data(pheno_df)

   GetResiduals(
     dnam = betasChr22_df[1:10, 1:10],
     betaToM = TRUE,
     pheno_df = pheno_df,
     covariates_char = c("age.brain", "sex", "slide")
   )

TransBioInfoLab/coMethDMR documentation built on Oct. 15, 2024, 12:52 a.m.