R/get_comparisons.R

In pmartR/pmartR: Panomics Marketplace - Quality Control and Statistical Analysis for Panomics Data

#' Returns data frame with comparisons to be made
#'
#' The method creates a data frame containing the comparisons to be made when
#' performing differential statistics.
#'
#' @param comp_type string specifying either "control" or
#'   "pairwise". Specifying "control" indicates that all other groups are
#'   to be compared to a single control group. Specifying "pairwise"
#'   indicates that all pairwise comparisons are to be made.
#' @param omicsData A pmartR data object of any class, which has a `group_df`
#'   attribute created by the `group_designation()` function
#' @param control_group string indicating the group to use for the control
#'   group. Only required when comp_type="control".
#'
#' @details This function takes in the omicsData and type of comparison, and
#'   returns a data frame where each row corresponds to a comparison of
#'   interest.
#'
#' @return data frame with columns for Test and Control. Each row corresponds to
#'   a comparison of interest.
#'
#' @examplesIf requireNamespace("pmartRdata", quietly = TRUE)
#' library(pmartRdata)
#' mymetab <- group_designation(omicsData = metab_object, main_effects = "Phenotype")
#' create_comparisonDF(comp_type = "pairwise", omicsData = mymetab)
#'
#' create_comparisonDF(comp_type = "control", omicsData = mymetab, control_group = "Phenotype1")
#'
#' @seealso \code{\link{group_designation}}
#'
#' @author Kelly Stratton
#'
#' @export
create_comparisonDF <- function(comp_type, omicsData, control_group = NULL) {
  # check that omicsData is of the appropriate class
  if (!inherits(omicsData, 
                c("proData", "pepData", "lipidData", "metabData", 
                  "nmrData", "seqData"))) {
    stop("omicsData is not an object of appropriate class")
  }

  # Check for group_DF attribute #
  if (is.null(attr(omicsData, "group_DF"))) {
    stop("group_designation must be called in order to create a 'group_DF' ",
         "attribute for omicsData.")
  } else {
    # group_data <- attr(omicsData, "group_DF")
    group_data <- get_group_DF(omicsData)
  }

  # do we even run this if they want custom groups?
  # need to address the order of the factor levels somewhere...
  #   maybe not in this function, but somewhere...


  ## initial checks ##
  if (!(comp_type %in% c("control", "pairwise"))) {
    stop("Comparison type must be one of the following: control, pairwise")
  }

  if (comp_type == "control" & is.null(control_group)) {
    stop("A control_group must be specified when comp_type is set to ",
         "'control'.")
  }

  # if(comp_type=="control" & (table(group_data$Group)[control_group]<2)){
  #  stop("The control_group selected has fewer than two subjects and cannot ",
  #       "be used.")
  # }

  ## end of initial checks ##

  grps <- unique(group_data$Group)
  n_grps <- length(grps)


  if (comp_type == "control") {
    query_grps <- grps[grps != control_group]
    control_grps <- rep(control_group, length(query_grps))
  }

  if (comp_type == "pairwise") {
    query_grps <- c()
    control_grps <- c()
    for (i in 1:n_grps) {
      for (j in (i + 1):n_grps) {
        if (i != j) {
          control_grps <- c(control_grps, as.character(grps[i]))
          query_grps <- c(query_grps, as.character(grps[j]))
        }
      }
    }
    # cbind(control_grps, query_grps)
    # remove any rows with an NA
    inds <- c(which(is.na(control_grps)), which(is.na(query_grps)))
    inds <- unique(inds)
    query_grps <- query_grps[-inds]
    control_grps <- control_grps[-inds]
  }


  comparisons <- data.frame(Test = query_grps, Control = control_grps)
  return(comparisons)
}