R/epi_mlm.R

In isglobal-brge/EpiMutations: Robust outlier identification for DNA methylation data

#' @title  Detects epimutations using Multivariate Linear Model (MLM)
#' @description  Identifies CpGs with outlier methylation values 
#' using methylated Multivariate Linear Model
#' @param mixture beta values matrix.  Samples in columns and
#' CpGs in rows. 
#' @param  model design (or model) matrix.
#' @return The function returns the F statistic, 
#' R2 test statistic and Pillai.
#' 
epi_mlm <- function(mixture, model) 
{
    mod <- mlm(t(mixture) ~ model[,2])
    statistics <- mod$aov.tab[1, c("F value", "R2", "Pr(>F)")]
    return(statistics)
}

#' @title  Creates a data frame containing the results 
#' obtained from MLM
#' @description Creates a data frame containing the
#' genomic regions, statistics and direction for the DMRs.
#' @param bump a DMR obtained from \link[bumphunter]{bumphunter}
#' (i.e. a row from \link[bumphunter]{bumphunter} method result).
#' @param sts the F statistic, R2 test statistic 
#' and Pillai obtained as a result
#' of \link[epimutacions]{epi_mlm} function. 
#' @returns The function returns a data 
#' frame containing the following 
#' information for each DMR: 
#' * genomic ranges
#' * DMR base pairs
#' * number and name of CpGs in DMR
#' * statistics: 
#'     * Outlier score
#'     * Outlier significance
#'     * Outlier direction
#'  * Sample name
#' 
#' For more information about the output see 
#' \link[epimutacions]{epimutations}.
#' 

res_mlm <- function(bump, sts) 
{
    
    bump$outlier_score <- paste0(sts[1], "/", sts[2])
    bump$outlier_direction <- ifelse(bump$value < 0, "hypomethylation",
                                                    "hypermethylation")
    bump$pvalue <- sts[3]
    bump$adj_pvalue <- NA
    bump[, c( "chromosome", "start", "end", "sz", "cpg_n", "cpg_ids",
        "outlier_score", "outlier_direction", "pvalue", "adj_pvalue",
        "delta_beta", "sample" )]
}