R/Subgraph.R

In idroz/gasub: R package for extraction of active subnetworks using a global search algorithm

#' Use genetic algorithm to identify maximally scoring subgraph
#' @param graph          a named igraph object
#' @param weights        a named numeric vector
#' @param pop.size       integer
#' @param max.iter       integer
#' @param run            integer
#' @param p.mutation     float. Probability of mutation
#' @param eletism        integer
#' @param ncores         integer
#'
#' @importFrom igraph get.edgelist vcount
#' @importFrom jsonlite toJSON
#' @export

Subgraph <- function(graph, weights, pop.size = 50, max.iter = 100, run = vcount(graph),
                    p.mutation = 0.1, eletism = max(1, round(pop.size * 0.05)), ncores = 1){

  # Error checking
  if (is.null (names(weights))) stop("Weights must be a named vector")
  if (is.null (V(graph)$name)) stop("Graph object must have a non-empthy V(graph)$name attribute")

  # Value checking
  if (p.mutation < 0 | p.mutation > 1) stop("Mutation probability must be 0-1")

  keep <- names(weights) %in% V(graph)$name
  weights <- weights[keep]

  if (length(weights)==0) stop("There are no matching names of the weight vector in the graph")

  # Set weights of missing nodes to 0
  missing <- setdiff(V(graph)$name, names(weights))
  m <- rep(0, length(missing))
  names(m) <- missing
  weights <- c(weights, m)

  # Create a one-one mapping between weight and vertex names
  ix <- match(V(graph)$name, names(weights))
  weights <- weights[ix]

  # Generate a json file using input options
  options <- list()
  options$edgelist <- get.edgelist(graph, names = FALSE)
  options$vertexlist <- V(graph)$name
  options$weights <- weights
  options$popsize <- pop.size
  options$pmutation <- p.mutation
  options$eletism <- eletism
  options$maxiter <- max.iter
  options$run <- run
  options$ncores <- ncores


  optionsfile <- paste0(getwd(), "/opts.json")
  popfile <- paste0(getwd(), "/population.csv")
  fitnessfile <- paste0(getwd(), "/fitness.csv")

  fileConn<-file(optionsfile)
  writeLines(toJSON(options, auto_unbox = TRUE), fileConn)
  close(fileConn)

  Genetic <- system.file("julia", "Subgraph.jl", package = "gasub")

  system(paste0("julia -p", ncores," ", Genetic, " ", optionsfile, " ", popfile, " ", fitnessfile))

  pop <- read.csv("population.csv", header = FALSE)
  fitness <- as.matrix(read.csv("fitness.csv", header = FALSE))

  unlink(optionsfile)
  unlink(popfile)
  unlink(fitnessfile)

  res <- structure(list(population = pop, fitness = fitness), class = "GA")
  return(res)

}