R/octadDrugEnrichment.R

In Bin-Chen-Lab/OCTAD: Open Cancer TherApeutic Discovery (OCTAD)

#' @export
#' @importFrom GSVA gsva
#' @importFrom limma barcodeplot
#' @import octad.db
#' @importFrom ExperimentHub ExperimentHub
#' @importFrom S4Vectors mcols
#' @importFrom AnnotationHub query
#' @importFrom octad.db get_ExperimentHub_data
#' @importFrom utils write.csv data txtProgressBar read.csv2 head read.csv
#' @importFrom grDevices pdf dev.off
#' 

octadDrugEnrichment <- function(sRGES = NULL, target_type = "chembl_targets", enrichFolder = "enrichFolder", outputFolder = NULL,
    outputRank = FALSE) {
    # require(GSVA)
    if (missing(sRGES)) {
        stop("sRGES input not found")
    }
    if (is.null(sRGES$sRGES) | is.null(sRGES$pert_iname)) {
        stop("Either sRGES or pert_iname collumn in Disease signature is missing")
    }
    options(warn = -1)

    if (is.null(outputFolder)) {
        outputFolder <- tempdir()
        message("outputFolder is NULL, writing output to tempdir()")
    }
    
    if(!dir.exists(outputFolder)){
      stop('Looks like output path ', outputFolder,' is either non-existent or obstructed. Check outputFolder option.')
    }

    if (!dir.exists(file.path(outputFolder, enrichFolder))) {
        dir.create(file.path(outputFolder, enrichFolder))
    }
    # eh_dataframe <- as.data.frame(S4Vectors::mcols(AnnotationHub::query(.eh, 'octad.db')))['title']
    # data('eh_dataframe',package='octad')
    eh_dataframe <- data.frame(title = c("cmpd_sets_ChemCluster", "cmpd_sets_chembl_targets", "cmpd_sets_mesh"))
    row.names(eh_dataframe) <- c("EH7266", "EH7267", "EH7268")


    random_gsea_score <- get_ExperimentHub_data("EH7275")
    for (target_type_selected in target_type) {
        message("Running enrichment for ", target_type_selected, "\n")
        enrichFolder.n <- file.path(outputFolder, enrichFolder, target_type_selected)
        if (!dir.exists(enrichFolder.n)) {
            dir.create(enrichFolder.n)
        }

        # load required random scores from octad.db
        eh_dataframe$object <- row.names(eh_dataframe)
        cmpd_sets <- get_ExperimentHub_data((eh_dataframe[eh_dataframe$title == paste0("cmpd_sets_", target_type_selected),
            "object"]))
        cmpdSets <- cmpd_sets$cmpd.sets
        names(cmpdSets) <- cmpd_sets$cmpd.set.names

        ############################


        drug_pred <- sRGES

        rgess <- matrix(-1 * drug_pred$sRGES, ncol = 1)

        if (is.null(dim(rgess))) {
            warning("rgess have zero rows, recompute it and try again")
            next
        } else {
            rownames(rgess) <- drug_pred$pert_iname
            rgess <- cbind(rgess, rgess)  # PLUG FOR GSVA BUG, FIX AS THEY WOULD FIX THEIR CODE
            gsea_results <- GSVA::gsva(rgess, cmpdSets, method = "ssgsea", parallel.sz = 8, ssgsea.norm = TRUE,
                verbose = FALSE)
            gsea_results <- gsea_results[-1, ]
            gsea_results <- merge(random_gsea_score[[target_type_selected]], gsea_results, by = "row.names")

            if (is.null(dim(gsea_results))) {
                warning("gsea_results have zero rows, recompute it and try again")
                next
            } else {
                rownames(gsea_results) <- gsea_results$Row.names
                gsea_results$Row.names <- NULL
                gsea_summary <- data.frame(score = gsea_results[, ncol(random_gsea_score[[target_type_selected]]) +
                  1])
                # calculating p.value
                gsea_p <- apply(gsea_results, 1, function(x) {
                  sum(x[seq_len(ncol(random_gsea_score[[target_type_selected]]))] > x[ncol(random_gsea_score[[target_type_selected]]) +
                    1])/ncol(random_gsea_score[[target_type_selected]])
                })
                gsea_p <- data.frame(target = names(gsea_p), score = gsea_summary, p = gsea_p, padj = p.adjust(gsea_p,
                  method = "fdr"))
                gsea_p <- gsea_p[order(gsea_p$padj), ]
                write.csv(gsea_p, file.path(enrichFolder.n, paste0("enriched_", target_type_selected, ".csv")),
                  row.names = FALSE)
                top.out.num <- nrow(gsea_p[which(gsea_p$padj <= 0.05), ])
                if (top.out.num == 0) {
                  top.out.num <- 1
                }
                if (top.out.num > 50) {
                  top.out.num <- 50
                }
                if (nrow(gsea_p) > 0) {
                  for (i in seq_len(top.out.num)) {
                    top_target <- as.character(gsea_p$target[i])
                    sRGES$rank <- rank(sRGES$sRGES)
                    target_drugs_score <- sRGES$rank[sRGES$pert_iname %in% cmpdSets[[top_target]]]
                    if (length(target_drugs_score) < 3) {
                      next
                    }
                    pdf(file.path(enrichFolder.n, paste0("/top_enriched_", top_target, "_", target_type_selected,
                      ".pdf")))
                    limma::barcodeplot(sRGES$sRGES, target_drugs_score, main = top_target, xlab = "sRGES")
                    dev.off()

                    # output detailed rank
                    if (outputRank) {
                      write.csv(sRGES[sRGES$pert_iname %in% cmpdSets[[top_target]], c("pert_iname", "rank")],
                        file.path(enrichFolder.n, paste0("/top_enriched_", top_target, "_", target_type_selected,
                          ".csv")))
                    }
                  }
                  if (target_type_selected == "ChemCluster") {
                    clusternames <- as.character((gsea_p[which(gsea_p$padj <= 0.05), ])$target)
                    if (length(clusternames) != 0) {
                      topclusterlist <- cmpdSets[clusternames]
                      #clusterdf <- read.csv2(paste0(enrichFolder.n, "misc.csv"), header = FALSE)
                      clusterdf=as.data.frame(sapply(topclusterlist, toString))
                      
                      clusterdf$cluster <- clusternames
                      clusterdf$pval <- (gsea_p[which(gsea_p$padj <= 0.05), ])$padj
                      colnames(clusterdf)[1] <- "drugs.in.cluster"
                      write.csv(clusterdf, file = file.path(enrichFolder.n, "drugstructureclusters.csv"), row.names = FALSE)
                    }
                  }
                  # msg <- paste('Done for', target_type_selected, 'for', nrow(gsea_p[which(gsea_p$padj <=
                  # 0.05), ]), 'genes')
                  message("Done for", target_type_selected, "for", nrow(gsea_p[which(gsea_p$padj <= 0.05), ]),
                    "genes")
                } else {
                  # msg <- paste('No signigicant enrichment found for', target_type_selected)
                  message("No signigicant enrichment found for", target_type_selected)
                }
            }
        }
    }
    options(warn = 0)
}