R/CalculateEnrichmentPValue.r

In CausalR: Causal network analysis methods

#' calculates an enrichment p-value
#' @description
#' Calculate a enrichment p-value for a given hypothesis by comparing the corresponding predicted and observed gene changes

#' @export
#' @concept CausalR
#' @param  predictions  predictions of changes from the CCG for a particular hypothesis
#' @param  results  gene changes observed in the experimental data
#' @return  an enrichment p-value
#' @examples
#' predictions <- matrix(c(1,2,3,1,1,-1), ncol = 2)
#' results<- matrix(c(1,2,3,4,1,1,-1,1), ncol = 2)
#' CalculateEnrichmentPValue(predictions, results)

CalculateEnrichmentPValue <- function(predictions, results) {
    
    setOfSignificantPredictions <- GetSetOfSignificantPredictions(predictions)
    setOfDifferentiallyExpressedGenes <- GetSetOfDifferentiallyExpressedGenes(results)
    setOfNonDifferentiallyExpressedGenes <- setdiff(results[, 1], setOfDifferentiallyExpressedGenes)
    
    # n_pp + n_pm + n_mp + n_mm
    numSignificantPredictionsThatAreResponsive <- length(intersect(setOfSignificantPredictions, setOfDifferentiallyExpressedGenes))
    # n+0 + n-0
    numSignificantPredictionsThatAreUnresponsive <- length(intersect(setOfSignificantPredictions, setOfNonDifferentiallyExpressedGenes))
    # n0+ + n0-
    numZeroPredictionsThatAreResponsive <- length(setOfDifferentiallyExpressedGenes) - numSignificantPredictionsThatAreResponsive
    # n00
    numZeroPredictionsThatAreUnreponsive <- length(setOfNonDifferentiallyExpressedGenes) - numSignificantPredictionsThatAreUnresponsive
    
    contingencyTable <- matrix(c(numSignificantPredictionsThatAreResponsive, numSignificantPredictionsThatAreUnresponsive, numZeroPredictionsThatAreResponsive, 
        numZeroPredictionsThatAreUnreponsive), nrow = 2)
    enrichmentPValue <- stats::fisher.test(contingencyTable, alternative = "greater")[[1]]
    
    return(enrichmentPValue)
}