survival.R

# MSomicsQC::surv_designation is a fucntion that creates a "surv_DF" attribute with elements:
#   1) t_death - defines column name in f_data
#   2) t_progress
#   3) ind_death
#   4) ind_progress
#   5) covariates
#
#   Must appear together:
#
#   t_death & ind_death (overall survival - OS option)
# OR
#   t_death, t_progress, ind_death AND ind_progress (progression free survival - PFS option)
#
#   covariates is always optional - two max for now
#
#   ---------
#   Once Lisa/Kelly makes the above, write function to create Kaplan-Meier Curves.
#   Only account for categorical variables when making these curves (separate curves for each level of each
#                                                                    covariate)
#
#   Functions skeleton:
#     plot_km <- function(omicsData,...Other figure options from pmartRqc plot functions...) - produce KM plot
#     summary_km <- function(omicsData, percent = ?) - return something like summary.survfit where "percent" gives a
#       survival prob and they want "time" variable returned


#' Basic survival analysis function
#'
#' Implements overall survival analysis or progression-free survival analysis, depending upon the datatypes supplied to
#' surv_designation, and return the "survfit" object
#'
#' @param omicsData A pmartR data object of any class, which has a `group_df` attribute that is usually created by the `group_designation()` function
#' @return if fitted survival analysis object is returned
#'
#' @examplesIf requireNamespace("pmartRdata", quietly = TRUE)
#' \dontrun{
#' library(MSomicsSTAT)
#' library(OvarianPepdataBP)
#'
#' # Basic analysis without covariates
#' attr(tcga_ovarian_pepdata_bp, "survDF") <- list(t_death = "survival_time",
#'                                                 ind_death = "vital_status")
#' sfit <- fit_surv(tcga_ovarian_pepdata_bp)
#' plot(sfit)
#'
#' # Add some covariate information
#' attr(tcga_ovarian_pepdata_bp, "survDF") <- list(
#'   t_death = "survival_time",
#'   ind_death = "vital_status",
#'   covariates = "g__initial_pathologic_diagnosis_method_g1"
#' )
#' sfit <- fit_surv(tcga_ovarian_pepdata_bp)
#' plot(sfit, col = c(1, 2))
#' }
#'
#' @export
fit_surv <- function(omicsData) {
  if (!requireNamespace("survival", quietly = TRUE)) {
    stop("Please install the 'survival' package.")
  }

  t_death <- attr(omicsData, "survDF")$t_death
  t_progress <- attr(omicsData, "survDF")$t_progress
  ind_death <- attr(omicsData, "survDF")$ind_death
  ind_progress <- attr(omicsData, "survDF")$ind_progress
  covariates <- attr(omicsData, "survDF")$covariates

  surv_df <- data.frame(time = omicsData$f_data[, t_death], status = 0)
  surv_df$status[grep("dead", omicsData$f_data[, ind_death], ignore.case = TRUE)] <- 1


  if (is.null(covariates)) {
    sfit <- survfit(Surv(time, status) ~ 1, data = surv_df)
  } else {
    form <- "Surv(time,status)~"
    ncovars <- length(covariates)

    for (i in 1:ncovars) {
      surv_df <- cbind(surv_df, omicsData$f_data[, covariates[i]])
      form <- paste0(form, covariates[i])
      if (i < ncovars) {
        form <- paste0(form, "+")
      }
    }
    colnames(surv_df)[ncol(surv_df) - c((ncovars - 1):0)] <- covariates

    sfit <- survfit(as.formula(form), data = surv_df)
  }
  return(sfit)
}


#' Basic survival analysis plot
#'
#' Implements overall survival analysis or progression-free survival analysis, depending upon the datatypes supplied to
#' surv_designation, and plot the resulting Kaplan-Meier curve.
#'
#' @param omicsData A pmartR data object of any class, which has a `group_df` attribute that is usually created by the `group_designation()` function
#'
#' @return a Kaplan-Meier curve
#'
#' @examplesIf requireNamespace("pmartRdata", quietly = TRUE)
#' \dontrun{
#' library(MSomicsSTAT)
#' library(OvarianPepdataBP)
#' attr(tcga_ovarian_pepdata_bp, "survDF") <- list(t_death = "survival_time",
#'                                                 ind_death = "vital_status")
#' plot_km(omicsData = tcga_ovarian_pepdata_bp)
#'
#' # Add covariates to "survDF" attribute
#' attr(tcga_ovarian_pepdata_bp, "survDF") <- list(
#'   t_death = "survival_time",
#'   ind_death = "vital_status",
#'   covariates = "age_at_initial_pathologic_diagnosis"
#' )
#' plot_km(omicsData = tcga_ovarian_pepdata_bp)
#' }
#'
#' @export
#'
plot_km <- function(omicsData) {
  sfit <- fit_surv(omicsData)


  plot(sfit)
}

#' Basic survival analysis summary
#'
#' Implements overall survival analysis or progression-free survival analysis, depending upon the datatypes supplied to
#' surv_designation, and gives a summary of the results.
#'
#' @param omicsData A pmartR data object of any class, which has a `group_df` attribute that is usually created by the `group_designation()` function
#' @param percent The percentile
#' @param ... extra arguments passed to regexpr if pattern is specified
#' @return if `percent` is provided then the time at which that probability of death is returned; else, the summary of the `survival` object is returned
#'
#' @examplesIf requireNamespace("pmartRdata", quietly = TRUE)
#' \dontrun{
#' library(OvarianPepdataBP)
#' attr(tcga_ovarian_pepdata_bp, "survDF") <- list(t_death = "survival_time",
#'                                                 ind_death = "vital_status")
#' # No percent is provided so the entire object is returned
#' summary_km(tcga_ovarian_pepdata_bp)
#'
#' # Percent is provided so corresponding time point is returned
#' summary_km(tcga_ovarian_pepdata_bp, .4)
#' }
#'
#' @export
#'
summary_km <- function(omicsData, percent = NULL, ...) {
  # Summary object
  sfit1 <- summary(fit_surv(omicsData, ...))

  if (!is.null(percent)) {
    row_of_interest <- which.min(abs(sfit1$surv - percent))
    return(sfit1$time[row_of_interest])
  }

  return(sfit1)
}