R/generate_graph.R

In b2slab/diffusion: Diffusion scores on biological networks

#' Generate data.frame with default vertex attributes
#'
#' @return data.frame with default node class attributes
#'
#' @rdname default_graph_param
#'
#' @import igraph
.default_graph_param <- function(){
    data.frame(
        color = c("#E64B35FF", "#4DBBD5FF", "#00A087FF"),
        shape = "circle",
        frame.color = "gray20",
        label.color = "gray5",
        size = 12,
        stringsAsFactors = FALSE,
        row.names = c("source", "filler", "end")
    )
}

#' Default proportions for randomly generated graphs
#'
#' @return named numeric with default class proportions
#'
#' @rdname default_graph_param
.default_prop <- c(source = .05, filler = .45, end = .5)

#' Function to connect a non connected graph
#'
#' @param g an igraph object
#'
#' @return a connected igraph object
#' @examples
#' library(igraph)
#' g <- diffuStats:::.connect_undirected_graph(
#'     graph.empty(10, directed = FALSE))
#' g
#' @rdname connect_undirected_graph
#'
#' @import igraph
.connect_undirected_graph <- function(g) {
    g.clusters <- clusters(g)

    # Partition nodes: in/out largest cc
    nodes.largest.cc <- which(
        g.clusters$membership == which.max(g.clusters$csize))
    nodes.addEdges <- which(
        g.clusters$membership != which.max(g.clusters$csize))

    if (length(nodes.addEdges) == 0) return(g)

    # For each node out, add one edge to a random node in
    g.newEdges <- vapply(
        nodes.addEdges,
        function(dummy) sample(nodes.largest.cc, 1), 
        1L)

    g <- add.edges(
        graph = g,
        edges = rbind(V(g)[nodes.addEdges], V(g)[g.newEdges]),
        directed = FALSE)

    g
}

#' Generate a random graph
#'
#' Function \code{generate_graph} generates a random network
#' using \pkg{igraph} graph generators. Several models are
#' available, and
#'
#' @param fun_gen function to generate the graphs. Typically from
#' \pkg{igraph}, like \code{\link[igraph:sample_pa]{barabasi.game}},
#' \code{\link[igraph:sample_smallworld]{watts.strogatz.game}},
#' \code{\link[igraph]{erdos.renyi.game}},
#' \code{\link[igraph]{make_lattice}}, etc.
#' @param param_gen list with parameters to pass to \code{fun_gen}
#' @param class_label character vector with length equal to
#' the number of nodes in the graph to generate.
#' If left to \code{NULL}, the default classes  are
#' \code{c("source", "filler", "end")} with proportions of
#' \code{c(0.05, 0.45, 0.5)}.
#' @param class_attr data.frame with vertex classes as rownames and a column
#' for each vertex attribute. The name of the column will be used as the
#' attribute name.
#' @param fun_curate function to apply to the graph before returning it.
#' Can be set to \code{identity} or \code{NULL} to skip this step.
#' By default, the graph is connected: nodes not belonging to the
#' largest connected component are randomly wired to a node in it.
#' @param seed numeric, seed for random number generator
#'
#' @return An \pkg{igraph} object
#'
#' @examples
#' g <- generate_graph(
#'     fun_gen = igraph::barabasi.game,
#'     param_gen = list(n = 100, m = 3, directed = FALSE),
#'     seed = 1)
#' g
#'
#' @import igraph
#' @export
generate_graph <- function(
    fun_gen,
    param_gen,
    class_label = NULL,
    class_attr = .default_graph_param(),
    fun_curate = .connect_undirected_graph,
    seed = NULL
) {
    if (!is.null(seed)) set.seed(seed)

    # Generate network using provided function
    g <- do.call(fun_gen, param_gen)
    n <- vcount(g)
    V(g)$name <- paste0("V", seq_len(n))

    # First assign classes
    if (is.null(class_label)) {
        message("Using default class proportions...")
        class <- rep(names(.default_prop), times = .default_prop*n)
        last_prop <- utils::tail(names(.default_prop), 1)

        # If rounding left the classes vector a bit short... complete it
        class <- c(class, rep(last_prop, n - length(class)))

        V(g)$class <- class
    } else {
        if (length(class_label) != n) {
            stop(
                "'class_label' must have its length ",
                "equalling the number of nodes.")
        }
        V(g)$class <- class_label
    }

    # Assign vertex attributes through a data_frame and vertex classes
    for (col in names(class_attr)) {
        mapper <- stats::setNames(class_attr[[col]], rownames(class_attr))
        g <- set_vertex_attr(
            g,
            name = col,
            value = mapper[V(g)$class])
    }

    # Connect graph (optional)
    if (is.null(fun_curate)) return(g)
    else return(do.call(fun_curate, list(g)))
}