R/topNet.R

In RegEnrich: Gene regulator enrichment analysis

# Filter the top edges in the network.
# @description Choose the top edges in the reulator-target network.
#
# @param network A data.frame of set (regulators),
# element (target genes) and weight.
# @param percent Numeric. The percentage (<= 10) of top edges, default is 5,
# which means top 5\% of Reg-target network edges will be retained.
# @param directed The type of the network: directed (\code{directed = TRUE})
# and undirected
# (\code{directed = FALSE}). Default is \code{TRUE}.
# @param reg The gene regulators. The default is \code{NULL}.
# @return A list of cutWeight (the cutoff edge weight),
# net (the genes connected by
# regulators), and validReg (valid regulators).
#' @rawNamespace import(dplyr, except = c(combine))
# @export
topNet = function(network, percent = 5, directed = TRUE, reg = NULL) {
  stopifnot(ncol(network) != 3 || is.data.frame(network))
  colnames(network) = c("from.gene", "to.gene", "weight")
  # stopifnot(identical(colnames(network), c("from.gene", "to.gene", 
  #                                          "weight")))
  stopifnot(percent <= 10)
  
  if (is.data.frame(network)) {
    network = as_tibble(network)
  }
  
  # from.gene are only from reg
  if (!is.null(reg)) {
    # network = network[network$from.gene %in% reg]
    network = filter(network, from.gene %in% reg)
  } else {
    reg = unique(as.character(network$from.gene))
  }
  
  if (!directed) {
    # reverse from.gene and to.gene
    networkRev = network
    networkRev$from.gene = network$to.gene
    networkRev$to.gene = network$from.gene
    # networkRev = network[network$from.gene %in% reg]
    networkRev = filter(network, from.gene %in% reg)
    
    # combine edges of two directions
    network = rbind(network, networkRev)
  }
  
  # networkReg = subset(network, !duplicated(network))
  networkReg = filter(network, !duplicated(network))
  n = nrow(networkReg)
  if (n == 0) {
    stop("No edges were selected! Please check from.gene and reg")
  }
  
  # top percent % edges of the network
  cutWeight = stats::quantile(networkReg$weight, 1 - percent/100)
  # networkRegTop = networkReg[networkReg$weight >= cutWeight]
  networkRegTop = filter(networkReg, weight >= cutWeight)
  
  reg = as.character(unique(networkRegTop$from.gene))
  networkRegTop$from.gene = as.character(networkRegTop$from.gene)
  networkRegTop$to.gene = as.character(networkRegTop$to.gene)
  
  # netTable = subset(networkRegTop, networkRegTop$from.gene %in% reg)
  netTable = filter(networkRegTop, from.gene %in% reg)
  colnames(netTable) = c("set", "element", "weight")
  
  res = list(netTable = netTable)
  return(res)
}