biopipen 0.29.1__py3-none-any.whl → 0.30.0__py3-none-any.whl

biopipen/scripts/scrna/SeuratPreparing-doublet_detection.R

@@ -0,0 +1,201 @@
+.get_envs_cached_doubletfinder <- function() {
+    envs_cache <- envs
+    envs_cache$ncores <- NULL
+    envs_cache$doublet_detector <- NULL
+    envs_cache$scDblFinder <- NULL
+    envs_cache$DoubletFinder$ncores <- NULL
+    envs_cache
+}
+
+.get_envs_cached_scdblfinder <- function() {
+    envs_cache <- envs
+    envs_cache$ncores <- NULL
+    envs_cache$doublet_detector <- NULL
+    envs_cache$DoubletFinder <- NULL
+    envs_cache$scDblFinder$ncores <- NULL
+    envs_cache
+}
+
+.run_doubletfinder <- function() {
+    library(DoubletFinder)
+    log_info("- Preparing Seurat object ...")
+
+    if (is.null(envs$DoubletFinder$ncores)) {
+        envs$DoubletFinder$ncores <- envs$ncores
+    }
+
+    # More controls from envs?
+    sobj <- FindNeighbors(sobj, dims = 1:envs$DoubletFinder$PCs)
+    sobj <- FindClusters(sobj)
+
+    log_info("- pK Indentification ...")
+    sweep.res.list <- paramSweep(
+        sobj,
+        PCs = 1:envs$DoubletFinder$PCs,
+        sct = envs$use_sct,
+        num.cores = envs$DoubletFinder$ncores
+    )
+    sweep.stats <- summarizeSweep(sweep.res.list, GT = FALSE)
+    bcmvn <- find.pK(sweep.stats)
+    bcmvn$Selected <- bcmvn$pK == bcmvn$pK[which.max(bcmvn$BCmetric)[1]]
+
+    pK <- bcmvn$pK[which.max(bcmvn$BCmetric)[1]]
+    pK <- as.numeric(as.character(pK))
+    pN <- envs$DoubletFinder$pN
+    log_info("- Homotypic Doublet Proportion Estimate ...")
+    homotypic.prop <- modelHomotypic(Idents(sobj))
+    nExp_poi <- round(nrow(sobj@meta.data) * envs$DoubletFinder$doublets)
+    nExp_poi.adj <- round(nExp_poi * (1 - homotypic.prop))
+
+    log_info("- Running DoubletFinder ...")
+    sobj <- doubletFinder(
+        sobj,
+        PCs = 1:envs$DoubletFinder$PCs,
+        pN = pN,
+        pK = pK,
+        nExp = nExp_poi.adj,
+        reuse.pANN = FALSE,
+        sct = envs$use_sct
+    )
+    pANN_col <- paste0("pANN_", pN, "_", pK)
+    pANN_col <- colnames(sobj@meta.data)[grepl(pANN_col, colnames(sobj@meta.data))]
+    DF_col <- paste0("DF.classifications_", pN, "_", pK)
+    DF_col <- colnames(sobj@meta.data)[grepl(DF_col, colnames(sobj@meta.data))]
+    doublets <- sobj@meta.data[, c(pANN_col, DF_col), drop = FALSE]
+    colnames(doublets) <-  c("DoubletFinder_score","DoubletFinder_DropletType")
+    doublets$DoubletFinder_DropletType <- tolower(doublets$DoubletFinder_DropletType)
+
+    pk_plot <- ggplot(bcmvn, aes(x = pK, y = BCmetric, color = Selected)) +
+        geom_point() +
+        # rotate x axis labels
+        theme(axis.text.x = element_text(angle = 90, hjust = 1))
+    list(doublets = doublets, pk_plot = pk_plot)
+}
+
+.run_scdblfinder <- function() {
+    library(scDblFinder)
+    if (is.null(envs$scDblFinder$ncores)) {
+        envs$scDblFinder$ncores <- envs$ncores
+    }
+
+    envs$scDblFinder$sce <- GetAssayData(sobj, layer = "counts")
+    if (envs$scDblFinder$ncores > 1) {
+        envs$scDblFinder$BPPARAM <- BiocParallel::MulticoreParam(envs$scDblFinder$ncores, RNGseed = 8525)
+    }
+    envs$scDblFinder$returnType <- "table"
+    envs$scDblFinder$ncores <- NULL
+
+    doublets <- do_call(scDblFinder, envs$scDblFinder)
+    doublets <- doublets[doublets$type == "real", , drop = FALSE]
+    doublets <- doublets[, c("score", "class"), drop = FALSE]
+    colnames(doublets) <- c("scDblFinder_score", "scDblFinder_DropletType")
+
+    list(doublets = doublets)
+}
+
+run_dd <- function(detector) {
+    log_info("Running {detector} ...")
+    if (detector == "DoubletFinder") {
+        envs_cache_fun <- .get_envs_cached_doubletfinder
+        run_fun <- .run_doubletfinder
+    } else if (detector == "scDblFinder") {
+        envs_cache_fun <- .get_envs_cached_scdblfinder
+        run_fun <- .run_scdblfinder
+    } else {
+        stop("Unknown doublet detector: ", detector)
+    }
+
+    cached <- get_cached(envs_cache_fun(), detector, cache_dir)
+    if (!is.null(cached$data)) {
+        log_info("- Loading cached results ...")
+        results <- cached$data
+    } else {
+        results <- run_fun()
+
+        cached$data <- results
+        save_to_cache(cached, detector, cache_dir)
+    }
+
+    results
+}
+
+save_dd <- function(dd, detector) {
+    doublets <- dd$doublets
+    write.table(
+        doublets,
+        file.path(joboutdir, paste0(detector, "_doublets_singlets.txt")),
+        row.names = FALSE,
+        quote = FALSE,
+        sep = "\t"
+    )
+
+    summary <- as.data.frame(table(dd$doublets[[paste0(detector, "_DropletType")]]))
+    colnames(summary) <- c("Classification", "Droplet_N")
+    write.table(
+        summary,
+        file.path(joboutdir, paste0(detector, "_summary.txt")),
+        row.names = FALSE,
+        quote = FALSE,
+        sep = "\t"
+    )
+
+    n_doublet <- summary$Droplet_N[summary$Classification == 'doublet']
+    log_info("- {n_doublet}/{sum(summary$Droplet_N)} doublets detected.")
+}
+
+add_dd_to_seurat <- function(sobj, dd) {
+    AddMetaData(sobj, metadata = as.data.frame(dd$doublets))
+}
+
+plot_dd <- function(sobj, dd, detector) {
+    if (detector == "DoubletFinder") {
+        log_debug("- Plotting pK vs BCmetric ...")
+        ggsave(dd$pk_plot, filename = file.path(plotsdir, "DoubletFinder_pK_BCmetric.png"))
+    }
+
+    log_info("- Plotting dimension reduction ...")
+    dimp <- DimPlot(
+        sobj, group.by = paste0(detector, "_DropletType"), order = "doublet",
+        cols = c("#333333", "#FF3333"), pt.size = 0.8, alpha = 0.5)
+    ggsave(dimp, filename = file.path(plotsdir, paste0(detector, "_dimplot.png")))
+}
+
+filter_dd <- function(sobj, dd, detector) {
+    subset(sobj,
+        cells = rownames(dd$doublets[
+            dd$doublets[[paste0(detector, "_DropletType")]] == "singlet", ,
+            drop = FALSE
+        ]))
+}
+
+report_dd <- function(detector) {
+    add_report(
+        list(
+            kind = "descr",
+            content = "The table contains the number of cells classified as singlets and doublets."
+        ),
+        list(
+            kind = "table",
+            data = list(path = file.path(joboutdir, paste0(detector, "_summary.txt")))
+        ),
+        h1 = paste0(detector, " Results"),
+        h2 = paste0("The ", detector, " Summary")
+    )
+
+    if (detector == "DoubletFinder") {
+        add_report(
+            list(name = "pK vs BCmetric", src = file.path(plotsdir, "pK_BCmetric.png")),
+            list(name = "Dimension Reduction Plot", src = file.path(plotsdir, "DoubletFinder_dimplot.png")),
+            ui = "table_of_images",
+            h1 = "DoubletFinder Results",
+            h2 = "Plots"
+        )
+    } else {
+        add_report(
+            list(name = "Dimension Reduction Plot",src = file.path(plotsdir, "scDblFinder_dimplot.png")),
+            ui = "table_of_images",
+            h1 = "scDblFinder Results",
+            h2 = "Plots"
+        )
+    }
+}