scRNA-seq_analysis

pipelines/16_train_doublets_classifier/apply_doublet_classifier.R

@@ -0,0 +1,80 @@
+args = commandArgs(trailingOnly=T)
+args = paste(args, collapse = "")
+args = unlist(strsplit(args, ";"))
+
+arguments.list = "
+seurat.addr.arg = args[1]
+save.at.arg     = args[2]
+doublet.svm.arg = args[3]
+"
+{
+expected_arguments = unlist(strsplit(arguments.list, "\n"))
+expected_arguments = expected_arguments[!(expected_arguments == "")]
+
+if(length(args) != length(expected_arguments)){
+  error.msg = sprintf('This pipeline requires %s parameters', as.character(length(expected_arguments)))
+  expected_arguments = paste(unlist(lapply(strsplit(expected_arguments, ".arg"), "[", 1)), collapse = "\n")
+  stop(sprintf('This pipeline requires these parameters: seurat.addr ; save.at (name of RDS file where processed data are saved), doublet.svm (folder name where singlet/doublet svm classifier for given organ is stored)'))
+}
+
+eval(parse(text = arguments.list))
+
+for(n in 1:length(expected_arguments)){
+  argument = expected_arguments[n]
+  argument = gsub(pattern=" ", replacement="", x=argument)
+  argument.name = unlist(strsplit(argument, "="))[1]
+  variable.name = gsub(pattern=".arg", replacement="", argument.name)
+  argument.content = eval(parse(text = argument.name))
+  eval(parse(text = argument.content))
+  if (!exists(variable.name)){
+    stop(sprintf("Argument %s not passed. Stopping ... ", variable.name))
+  }
+}
+
+# create required folders for output and work material
+output_folder = gsub(pattern="^\\d+_", replacement="", x=basename(getwd()))
+output_folder = paste(output_folder, seurat.addr, sep = "_")
+c.time = Sys.time()
+c.time = gsub(pattern=" BST", replacement="", x=c.time)
+c.time = gsub(pattern=":", replacement="", x=c.time)
+c.time = gsub(pattern=" ", replacement="", x=c.time)
+c.time = gsub(pattern="-", replacement="", x=c.time)
+c.time = substr(x=c.time, start=3, stop=nchar(c.time))
+output_folder = paste(output_folder, c.time, sep = "_")
+output_folder = file.path("../../output", output_folder)
+dir.create(output_folder)
+
+output_folder_material = file.path(output_folder, "material")
+dir.create(output_folder_material)
+
+seurat.addr = file.path("../../data", seurat.addr)
+
+source("../../tools/bunddle_utils.R")
+
+save.at = file.path("../../data", save.at)
+
+library(Seurat)
+library(RColorBrewer)
+library(dplyr)
+library(plyr)
+}
+###############################################################################
+# Load data
+print("Loading data ...")
+seurat.obj = readRDS(seurat.addr) # seurat.addr.arg
+print("Data loaded.")
+
+print("Identifying doublets")
+seurat.obj@meta.data$doublets <- Apply_Classifier_On_Seurat_Object(
+  seurat.obj = seurat.obj,
+  classifier.fname = doublet.svm, # doublet.svm.arg
+  tool_addr = tool_addr,
+  python.addr = python.addr)
+
+print("Doublets and singlets: ")
+print(table(seurat.obj@meta.data$lanes, seurat.obj@meta.data$doublets))
+
+print("Saving data")
+saveRDS(seurat.obj, save.at) # save.at.arg
+
+unlink(output_folder_material, recursive = T, force = T)

pipelines/16_train_doublets_classifier/apply_doublet_classifier.sh

@@ -0,0 +1,16 @@
+#!/bin/bash
+
+#$ -cwd
+#$ -N apply_doublets_SVM
+#$ -V
+#$ -l h_rt=23:59:59
+#$ -l h_vmem=100G
+
+if [ "$#" -ne 1 ]; then
+    echo "Illegal number of parameters"
+    exit 1
+fi
+
+Rscript apply_doublet_classifier.R $1
+
+echo "End on `date`"

pipelines/16_train_doublets_classifier/svm.py

@@ -0,0 +1,54 @@
+import pandas as pd
+from sklearn.metrics import classification_report
+from sklearn.metrics import confusion_matrix
+from sklearn.svm import SVC
+from sklearn.model_selection import GridSearchCV
+import pickle
+
+import sys
+from os.path import join
+args = sys.argv
+save_to = args[1]
+
+print("Loading data ...")
+X = pd.read_csv(join(save_to, "./data.csv"), sep = ",", index_col = 0).values
+y = pd.read_csv(join(save_to, "./labels.csv")).values[:, 0].reshape(-1, 1).ravel()
+
+from sklearn.decomposition import PCA
+pca = PCA(n_components = .8)
+X = pca.fit_transform(X)
+
+modelFile = open(join(save_to, "pca.pickle"), "wb")
+print(modelFile)
+modelFile.write(pickle.dumps(pca))
+modelFile.close()
+
+print(X.shape)
+
+from sklearn.model_selection import train_test_split
+X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.33, random_state=19)
+
+params = {"C":[1, 10, 100, 300]}
+
+print("Creating the model and fitting the data ...")
+model = GridSearchCV(SVC(probability = False, kernel = "rbf"), params, cv=3)
+model.fit(X_train, y_train)
+
+print("Testing ...")
+pred = model.predict(X)
+cls_report = classification_report(y, pred, target_names = model.classes_)
+print(cls_report)
+with open(join(save_to, "classification_report.txt"), "w")  as cl_f:
+    cl_f.write(cls_report)
+
+print("Saving model and confusion matrix to disk ...")
+cnf_matrix = confusion_matrix(y, pred)
+df = pd.DataFrame(cnf_matrix)
+df.columns = model.classes_
+df.to_csv(join(save_to, "confusion_matrix.csv"))
+
+modelFile = open(join(save_to, "model.pickle"), "wb")
+modelFile.write(pickle.dumps(model))
+modelFile.close()
+
+print(model.best_params_)

pipelines/16_train_doublets_classifier/train_doublets_SVM.R

@@ -0,0 +1,119 @@
+args = commandArgs(trailingOnly=T)
+args = paste(args, collapse = "")
+args = unlist(strsplit(args, ";"))
+
+arguments.list = "
+seurat.addr.arg = args[1]
+save.to.arg     = args[2]
+"
+
+expected_arguments = unlist(strsplit(arguments.list, "\n"))
+expected_arguments = expected_arguments[!(expected_arguments == "")]
+
+if(length(args) != length(expected_arguments)){
+  error.msg = sprintf('This pipeline requires %s parameters', as.character(length(expected_arguments)))
+  expected_arguments = paste(unlist(lapply(strsplit(expected_arguments, ".arg"), "[", 1)), collapse = "\n")
+  stop(sprintf('This pipeline requires %s parameters: '))
+}
+
+eval(parse(text = arguments.list))
+
+for(n in 1:length(expected_arguments)){
+  argument = expected_arguments[n]
+  argument = gsub(pattern=" ", replacement="", x=argument)
+  argument.name = unlist(strsplit(argument, "="))[1]
+  variable.name = gsub(pattern=".arg", replacement="", argument.name)
+  argument.content = eval(parse(text = argument.name))
+  eval(parse(text = argument.content))
+  if (!exists(variable.name)){
+    stop(sprintf("Argument %s not passed. Stopping ... ", variable.name))
+  }
+}
+
+seurat.addr = file.path("../../data", seurat.addr)
+
+source("../../tools/bunddle_utils.R")
+
+library(Seurat)
+library(RColorBrewer)
+library(plyr)
+library(dplyr)
+
+#######################################################################################################
+
+# load data
+print("loading data ... ")
+seurat.obj = readRDS(seurat.addr)
+print("Data loaded.")
+
+save.to = file.path("../../resources", save.to)
+
+dir.create(save.to)
+
+# for HdCA samples:
+#seurat.obj@meta.data$lanes <- seurat.obj@meta.data$biomaterial_id
+
+# mix cells within lanes
+lanes <- as.vector(unique(seurat.obj@meta.data$lanes))
+mixersA <- c(); mixersB <- c();
+for (i in 1:length(lanes)){
+  lane = lanes[i]
+  print(lane)
+  lane.indices <- which(seurat.obj@meta.data$lanes == lane)
+  mixersA <- c(mixersA, sample(x=lane.indices, size=length(lane.indices), replace=T))
+  mixersB <- c(mixersB, sample(x=lane.indices, size=length(lane.indices), replace=T))
+}
+
+# create a matrix of raw data for singlets - these are the cell collected
+singlets <- seurat.obj@raw.data[rownames(seurat.obj@data), colnames(seurat.obj@data)]
+
+# create doublets: 1) select from the singlets matrix 2 other matrices using the mixersA and mixersB; 2) add these 2 matrices
+doublets <-  singlets[, mixersA] + singlets[, mixersB]
+
+# make the cell names unique
+colnames(singlets) <- paste("Singlet",  1:dim(singlets)[2], sep = "_")
+colnames(doublets) <- paste("Doublets", 1:dim(doublets)[2], sep = "_")
+
+# merge the singlets and doublets sparse matrices
+# need to look into whether cBind deprecation will affect results
+merged.matrix <- Matrix::cBind(singlets, doublets)
+
+# free up some space
+rm(seurat.obj, lanes, mixersA, mixersB)
+
+# create label vector
+labels <- rep(c("Singlet", "Doublet"), each = ncol(singlets))
+labels <- data.frame(Labels = labels)
+
+# create a seurat object from singlets and doublets
+seurat.obj <- CreateSeuratObject(raw.data = merged.matrix, project="Doublet_Classifier", min.cells=0, min.genes=0)
+
+print("Checkpoint 2")
+# free up some space
+rm(singlets, doublets, merged.matrix)
+print("Checkpoint 3")
+# normalize the data
+seurat.obj <- NormalizeData(object = seurat.obj, normalization.method = "LogNormalize", scale.factor = 10000)
+# compute variable genes
+seurat.obj <- FindVariableGenes(object = seurat.obj, mean.function = ExpMean, 
+                                dispersion.function = LogVMR, x.low.cutoff = .0125, 
+                                x.high.cutoff = 3, y.cutoff = .625)
+# save variable genes in the same folder as the classifier
+classifier.features <- seurat.obj@var.genes
+saveRDS(classifier.features, file.path(save.to, "feature_genes.RDS"))
+print("Checkpoint 4")
+
+shuffle.key <- sample(x=1:dim(labels)[1], size=dim(labels)[1], replace=F)
+
+# save the labels and normalized data to disk 
+write.csv(labels[shuffle.key,], file.path(save.to, "labels.csv"), row.names = F)
+x.data <- as.data.frame(t(as.matrix(seurat.obj@data[classifier.features, ])))[shuffle.key,]
+print(dim(x.data))
+write.csv(x.data, file.path(save.to, "data.csv"), row.names = T)
+
+command = sprintf("%s svm.py %s", python.addr, save.to)
+system(command, wait = T)
+
+file.remove(file.path(save.to, "data.csv"), file.path(save.to, "labels.csv"), './Rplots.pdf')
+
+print("Ended beautifully ... ")

pipelines/16_train_doublets_classifier/train_doublets_SVM.sh

@@ -0,0 +1,16 @@
+#!/bin/bash
+
+#$ -cwd
+#$ -N train_doublets_SVM
+#$ -V
+#$ -l h_rt=47:59:59
+#$ -l h_vmem=300G
+
+if [ "$#" -ne 1 ]; then
+    echo "Illegal number of parameters"
+    exit 1
+fi
+
+Rscript train_doublets_SVM.R $1
+
+echo "End on `date`"