R/create_SiFINeT_object.R

In SiFINeT: Single Cell Feature Identification with Network Topology

#' The SiFINeT Class
#'
#' @slot data a list of cell (row) by gene (column) count matrix, either regular or sparse matrix
#' @slot sparse whether the count matrix should be analyzed as sparse matrix
#' @slot meta.data matrix of meta data, the number of rows should equal to the number of cells
#' @slot gene.name a vector of names of genes with length equal to the number of genes
#' @slot data.name name of the dataset
#' @slot n number of cells in the dataset
#' @slot p number of genes in the dataset
#' @slot data.thres binarized count matrix
#' @slot coexp matrix of genes coexpression
#' @slot est_ms estimated mean and sd of coexpression values
#' @slot thres lower bound of coexpression (or absolute value of coexpression) for network edge assignment
#' @slot q5 50% quantile for each gene
#' @slot kset index of kept genes after the filtering step
#' @slot conn list of connectivities in absolute network
#' @slot conn2 list of connectivities in positive sub-network
#' @slot fg_id index of the candidate feature genes
#' @slot uni_fg_id index of the candidate unique feature genes
#' @slot uni_cluster cluster result of the candidate unique feature genes
#' @slot selected_cluster selected unique feature gene clusters
#' @slot featureset detected set of feature genes
#'
#' @name SiFINeT-class
#' @exportClass SiFINeT
#'
SiFINeT <- setClass(
  Class = "SiFINeT",
  slots = c(
    data = "list",
    sparse = "logical",
    meta.data = "matrix",
    gene.name = "vector",
    data.name = "character",
    n = "numeric",
    p = "numeric",
    data.thres = "list",
    coexp = "matrix",
    est_ms = "list",
    thres = "numeric",
    q5 = "numeric",
    kset = "integer",
    conn = "list",
    conn2 = "list",
    fg_id = "integer",
    uni_fg_id = "integer",
    uni_cluster = "numeric",
    selected_cluster = "numeric",
    featureset = "list"
  )
)

#' create_SiFINeT_object
#'
#' The function classifies count data based on thresholds
#' defined by quantile regression
#' @param counts count matrix
#' @param gene.name name of the features
#' @param meta.data data.frame of meta data
#' @param data.name name of dataset
#' @param sparse whether the count matrix should be analyzed as sparse matrix
#' @param rowfeature whether the count matrix is feature (row) by cell (column)
#' @return a SiFINeT object
#' @export
#'
create_SiFINeT_object <- function(counts, gene.name = NULL,
                                  meta.data = NULL, data.name = NULL,
                                  sparse = FALSE, rowfeature = TRUE) {
  if (rowfeature == TRUE) {
    counts <- t(counts)
  }
  if (is.null(gene.name)) {
    gene.name <- colnames(counts)
  }
  if (is.null(gene.name)) {
    gene.name <- 1:ncol(counts)
  }
  if (is.null(data.name)) {
    data.name <- "data1"
  }
  if (is.null(meta.data)) {
    meta.data <- matrix(0, nrow(counts), 0)
  }
  data <- list(counts)
  names(data) <- data.name
  object <- new(
    Class = "SiFINeT",
    data = data,
    sparse = sparse,
    meta.data = meta.data,
    gene.name = gene.name,
    data.name = data.name,
    n = nrow(counts),
    p = ncol(counts),
    q5 = apply(counts, 2, quantile, 0.5),
    kset = 1:ncol(counts),
    featureset = list(
      unique = list(),
      shared = list(),
      enriched = list()
    )
  )
  return(object)
}