Merge branch 'issue-26' into dev (close #26, #43)

DESCRIPTION

@@ -1,7 +1,7 @@
 Package: DIscBIO
 Date: 2021-04-28
 Title: A User-Friendly Pipeline for Biomarker Discovery in Single-Cell Transcriptomics
-Version: 1.2.0.9003
+Version: 1.2.0.9004
 Authors@R:
   c(
     person(
@@ -38,7 +38,7 @@ Description: An open, multi-algorithmic pipeline for easy, fast and efficient
   <doi:10.1101/700989>.
 License: MIT + file LICENSE
 Encoding: UTF-8
-Imports: methods, TSCAN, boot, httr, mclust, statmod, igraph,
+Imports: methods, TSCAN, httr, mclust, statmod, igraph,
   RWeka, philentropy, NetIndices, png, grDevices,
   RColorBrewer, ggplot2, rpart, fpc, cluster, rpart.plot,
   tsne, AnnotationDbi, org.Hs.eg.db, graphics, stats, utils, impute,
@@ -87,6 +87,7 @@ Collate:
     'RpartDT.R'
     'RpartEVAL.R'
     'VolcanoPlot.R'
+    'bootstrap.R'
     'calc_pcareduceres.R'
     'cross.val.R'
     'customConverters.R'

NAMESPACE

@@ -46,7 +46,6 @@ importFrom(NetIndices,GenInd)
 importFrom(RColorBrewer,brewer.pal)
 importFrom(RWeka,J48)
 importFrom(TSCAN,TSCANorder)
-importFrom(boot,boot)
 importFrom(cluster,clusGap)
 importFrom(cluster,maxSE)
 importFrom(cluster,silhouette)

R/Jaccard.R

@@ -10,48 +10,26 @@
 #' @param K A numeric value of the number of clusters
 #' @param plot if `TRUE`, plots the mean Jaccard similarities
 #' @param R number of bootstrap replicates
-#' @param ... Further arguments passed to \code{boot::boot}
 #' @importFrom philentropy distance
-#' @importFrom boot boot
 #' @importFrom graphics barplot box
 #' @return A plot of the mean Jaccard similarity coefficient per cluster.
-Jaccard <- function(
-    object,
-    Clustering = "K-means",
-    K,
-    plot = TRUE,
-    R = 100,
-    ...) {
+Jaccard <- function(object, Clustering = "K-means", K, plot = TRUE, R = 100) {
   JACCARD <- vector()
 
   # Validation
   if (!(Clustering %in% c("K-means", "MB"))) {
     stop("Clustering has to be either K-means or MB")
   }
-
-  JS <- function(data, indices) {
-    d <- data[indices, ] # allows boot to select sample
-    jac <- suppressMessages(distance(t(d), method = "jaccard"))
-    jac1 <- 1 - jac
-    JSmean <- mean(jac1)
-    return(JSmean)
-  }
   for (i in 1:K) {
     # Optimize by avoiding if every loop. Only thing variable is data
     if (Clustering == "K-means") {
       target_col <- object@kmeans$kpart
     } else if (Clustering == "MB") {
       target_col <- object@MBclusters$clusterid
     }
-    results <- boot(
-      data      = object@fdata[, which(target_col == i)],
-      statistic = JS,
-      R         = R,
-      stype     = "f",
-      ...
-    )
+    results <- bootstrap(object@fdata[, which(target_col == i)], R)
     # to get the mean of all bootstrappings (mean of mean Jaccard values)
-    JACCARD[i] <- round(mean(results$t), digits = 3)
+    JACCARD[i] <- round(mean(results), digits = 3)
   }
   if (plot) {
     barplot(

R/bootstrap.R

@@ -0,0 +1,20 @@
+bootstrap <- function(data, n_reps) {
+  results <- numeric(n_reps)
+  n <- nrow(data)
+  i <- matrix(sample(n, n * n_reps, replace = TRUE), n_reps, n)
+  freqs <- t(apply(i, 1, tabulate, ncol(i)))
+  for (rep in seq_len(n_reps)) {
+    samp <- freqs[rep, ]
+    jac <- JS(data, samp)
+    results[rep] <- mean(jac)
+  }
+  return(results)
+}
+
+JS <- function(data, indices) {
+  d <- data[indices, ]
+  jac <- suppressMessages(distance(t(d), method = "jaccard"))
+  jac1 <- 1 - jac
+  JSmean <- mean(jac1)
+  return(JSmean)
+}

README.md

@@ -70,7 +70,7 @@ if (!requireNamespace("BiocManager", quietly = TRUE)) install.packages("BiocMana
 
 BiocManager::install(
     c(
-        "SingleCellExperimentmethods", "TSCAN", "boot", "httr", "mclust",
+        "SingleCellExperimentmethods", "TSCAN", "httr", "mclust",
         "statmod", "igraph", "RWeka", "philentropy", "NetIndices", "png",
         "grDevices", "RColorBrewer", "ggplot2", "rpart", "fpc",
         "cluster", "rpart.plot", "tsne", "AnnotationDbi", "org.Hs.eg.db",