R/generatePathwaysNetwork.R

In krumsiek/MoDentify: Phenotype-driven module identification

#' Subnetwork Creation function
#'
#'
#' @param data a \code{\link[data.table]{data.table}} or matrix containing data.
#' The columns correspond to the variables, the rows to the observations.
#' @param covars a \code{\link[data.table]{data.table}} containing covariates to correct for.
#' The columns correspond to the different covariates, the rows to the observations.
#' @param annotations  a \code{\link[data.table]{data.table}} containing annotations for the variables.
#' The columns correspond to the different annotations, the rows to the variables.
#' @param correlation.type type of correlation to be estimated. Can either be "pearson", or "partial".
#' @param level name of the column that should be used for grouping the variables. The default is "SUB_PATHWAY".
#' @param alpha significance threshold (type 1 error) for multiple testing correction.
#' @param correction.method the method that should be used for multiple testing correction ("bonferroni", "BH", "BY", "fdr", "holm", "hochberg", "hommel", "none").
#' Default is bonferroni. See \code{\link[stats]{p.adjust}}.
#' @param rm.unknown remove variables, for which the group is unknown. The corresponding grouping variable name should be "Unknown".
#' Default is \code{TRUE}.
#' @param representative.method the method, that is used for calculation of the eigenmetabolites.
#' Currently implemented: "eigenmetabolite" and "average"
#' @import igraph
#' @export
#' @examples
#' data(qmdiab.data)
#' data(qmdiab.annos)
#' 
#' pathway.graph <- generatePathwaysNetwork(
#'   data = qmdiab.data,
#'   annotations = qmdiab.annos, level = "Sub.pathway"
#' )
#' @return a list with a network containing the variables as nodes as
#' an \code{\link[igraph]{igraph}} object and a list with the module
#' representatives including their explained variances.
#' @references
#' \insertRef{Krumsiek2011}{MoDentify}
#' @references
#' \insertRef{Do2017}{MoDentify}
generatePathwaysNetwork <- function(data, covars = NULL, annotations,
                                    level = "Sub.pathway",
                                    correlation.type = "partial",
                                    alpha = 0.05,
                                    correction.method = "bonferroni", rm.unknown = TRUE,
                                    representative.method = "eigenmetabolite") {
  annotations <- as.data.table(annotations)
  annotations <- orderAnnotation(annotations, colnames(data))

  data <- scale(data)
  subannotations <- copy(annotations)
  subannotations[, label := get(level)]
  if (!"Fluid" %in% colnames(annotations)) {
    subannotations[, name := subannotations[, get(level)]]
  } else {
    subannotations[, name := paste(subannotations[, get(level)], Fluid)]
  }

  if (representative.method == "eigenmetabolite") {
    if (sum(is.na(data)) > 0) {
      stop("Data matrix contains missing values.\n Pathway network via eigenemtabolite approach not possible.\n")
    } else {
      eigenMetabolites <- eigenMetabolites(data = data, group = subannotations[, name], method = "PCA")
      mat <- as.matrix(eigenMetabolites$M)
      representatives <- eigenMetabolites
    }
  } else if (representative.method == "average") {
    representatives <- NULL
    mat <- as.matrix(aggregatedMean(data = data, group = subannotations[, name]))
    representatives$M <- mat
    representatives$expvar <- NULL
  } else {
    stop("\n Method not supported.\n")
  }


  subannotations <- subset(subannotations, !duplicated(name))

  if (rm.unknown) {
    subannotations <- subannotations[!grepl("Unknown|NA", get(level)), ]
    mat <- mat[, !grepl("Unknown|NA", colnames(mat))]
  }

  return(list(
    network = generateNetwork(mat, covars = covars, annotations = subannotations, correlation.type = correlation.type, alpha = alpha),
    representatives = representatives, level = level
  ))
}