R/computeDSS.R

In PharmacoGx: Analysis of Large-Scale Pharmacogenomic Data

##TODO:: Add function documentation
computeDSS <- function(concentration,
                       viability,
                       Hill_fit,
                       t_param = 10,
                       conc_as_log = FALSE,
                       viability_as_pct = TRUE,
                       trunc = TRUE,
                       verbose = TRUE,
                       dss_type = 3,
                       censor = FALSE
                       #, ...
) {
  
  if(missing(concentration)){
    stop("The concentration values to integrate over must always be provided.")
  }
  if (missing(Hill_fit)) {
    
    Hill_fit <- logLogisticRegression(concentration,
                                      viability,
                                      conc_as_log = conc_as_log,
                                      viability_as_pct = viability_as_pct,
                                      trunc = trunc,
                                      verbose = verbose)
    cleanData <- sanitizeInput(conc = concentration, 
                               Hill_fit = Hill_fit,
                               conc_as_log = conc_as_log,
                               viability_as_pct = viability_as_pct,
                               trunc = trunc,
                               verbose = verbose)
    pars <- cleanData[["Hill_fit"]]
    concentration <- cleanData[["log_conc"]]
    
  } else {
    
    cleanData <- sanitizeInput(conc = concentration, 
                               viability = viability,
                               Hill_fit = Hill_fit,
                               conc_as_log = conc_as_log,
                               viability_as_pct = viability_as_pct,
                               trunc = trunc,
                               verbose = verbose) #is this coercing the concentration to log?
    pars <- cleanData[["Hill_fit"]]
    concentration <- cleanData[["log_conc"]]
    
  }
  
  if(pars[[3]] > max(concentration)) {
    return(0)
  }
  
  if(!viability_as_pct){
    t_param = t_param * 100
    pars[[2]] <- pars[[2]] * 100
  }
  
  x2 = max(concentration)
  x1 = computeICn(concentration = concentration, Hill_fit = unlist(pars), n = t_param, conc_as_log = TRUE, viability_as_pct = TRUE)
  if(!is.finite(x1)){return(0)}
  
  x1 <- max(x1, min(concentration))
  
  if (censor) {
    if (pars[[2]] > 50) {
      return(NA)
    } else if (all(concentration < pars[[3]])) {
      return(0)
    }
  }
  
  AUC <- computeAUC(concentration = c(x1, x2), Hill_fit = unlist(pars), conc_as_log = TRUE, viability_as_pct = TRUE, verbose = verbose, trunc = trunc)
  
  
  DSS <- (AUC * (x2 - x1) - t_param * (x2 - x1)) / ((100 - t_param) * (max(concentration) - min(concentration)))
  if (dss_type == 1) {
    return(DSS)
  }
  DSS <- DSS / log(100 - pars[[2]])
  if (dss_type == 2) {
    return(DSS)
  }
  DSS <- DSS * (x2 - x1) / (max(concentration) - min(concentration))
  if (dss_type == 3) {
    return(DSS)
  } else {
    stop("Invalid DSS type entered.")
  }
}