scRNA-seq_analysis

pipelines/18_clustering_comparison/clustering.py

@@ -0,0 +1,91 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+Created on Wed Oct 03 10:49:22 2018
+
+@author: doru
+"""
+
+import sys
+args = sys.argv
+output_folder = args[1]
+n_clusters = int(args[2])
+
+import os
+
+os.chdir(output_folder)
+
+from sklearn.cluster import AgglomerativeClustering
+from sklearn.neighbors import kneighbors_graph
+from sklearn import mixture
+from sklearn import metrics
+import pandas as pd
+import numpy as np
+
+print("Loading data ...")
+pca_df = pd.read_csv("material/pca.csv", sep = ",", index_col = 0)
+pca = pca_df.values[:, :20]
+ori_labels = pca_df.values[:, 20]
+
+print("computing agglomerative clustering ... ")
+connectivity_graph = kneighbors_graph(X = pca, n_neighbors = 30)
+clustering = AgglomerativeClustering(n_clusters = n_clusters, affinity = "euclidean", 
+                                     connectivity = connectivity_graph, 
+                                     linkage = "ward")
+agg_labels = clustering.fit(pca)
+agg_labels = agg_labels.labels_
+
+print("computing gaussian mixture clustering ... ")
+gmm = mixture.GaussianMixture(n_components = n_clusters, covariance_type = "full",
+                              random_state = 52).fit(pca)
+gmm_labels = gmm.predict(pca)
+
+# metrics
+
+# voted labels
+def voted_labels(true, pred):
+    pred_unique = np.unique(pred)
+    pred_voted = true.copy()
+    for p in pred_unique:
+        p_indx = pred == p
+        ori_types = true[p_indx]
+        ori_unique = np.unique(ori_types, return_counts = True)
+        voted = ori_unique[0][np.argmax(ori_unique[1])]
+        pred_voted[p_indx] = voted
+    return pred_voted
+
+agg_labels = voted_labels(ori_labels, agg_labels)
+gmm_labels = voted_labels(ori_labels, gmm_labels)
+
+# compute adjusted Rand Index
+# reference: Comparing partitions - https://link.springer.com/article/10.1007%2FBF01908075
+RandIndex_Louvain_Agg = metrics.adjusted_rand_score(ori_labels, agg_labels)
+RandIndex_Louvain_GMM = metrics.adjusted_rand_score(ori_labels, gmm_labels)
+
+# ref for mutual information metrics - "Information theoretic measures for clusterings comparison"
+
+# compute adjusted mutual information
+AdjMutInf_Louvain_Agg = metrics.adjusted_mutual_info_score(ori_labels, agg_labels)
+AdjMutInf_Louvain_GMM = metrics.adjusted_mutual_info_score(ori_labels, gmm_labels)
+
+RandIndex_Louvain_Agg = str(RandIndex_Louvain_Agg)
+RandIndex_Louvain_GMM = str(RandIndex_Louvain_GMM)
+AdjMutInf_Louvain_Agg = str(AdjMutInf_Louvain_Agg)
+AdjMutInf_Louvain_GMM = str(AdjMutInf_Louvain_GMM)
+
+# save measures to disk
+result = '\n'.join([RandIndex_Louvain_Agg, RandIndex_Louvain_GMM, 
+                    AdjMutInf_Louvain_Agg, AdjMutInf_Louvain_GMM])
+
+with open('material/agreement_measures.txt', 'w') as agg_file:
+  agg_file.write(result)
+  
+print("saving gaussian mixture clustering labels to disk ...")
+df = pd.DataFrame(gmm_labels)
+df.index = pca_df.index
+df.to_csv("./material/gaussian_mixture.csv")
+
+print("saving agglomerative clustering labels to disk ...")
+df = pd.DataFrame(agg_labels)
+df.index = pca_df.index
+df.to_csv("./material/agglomerative_clustering.csv")

pipelines/18_clustering_comparison/clustering_comparison.R

@@ -0,0 +1,170 @@
+args = commandArgs(trailingOnly=T)
+args = paste(args, collapse = "")
+args = unlist(strsplit(args, ";"))
+
+arguments.list = "
+seurat.addr.arg    = args[1]
+set.ident.arg      = args[2]
+n_clusters.arg     = args[3]
+type.to.colour.arg = args[4]
+"
+
+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))
+  }
+}
+
+# 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)
+
+n_clusters = as.numeric(n_clusters)
+
+source("../../tools/bunddle_utils.R")
+
+library(Seurat)
+library(plyr)
+library(dplyr)
+library(reshape2)
+library(RColorBrewer)
+
+#######################################################################################################
+
+# load data
+print("loading data ... ")
+seurat.obj = readRDS(seurat.addr)
+print("Data loaded.")
+
+seurat.obj = SetAllIdent(seurat.obj, id=set.ident)
+
+# saving pca data to disk
+pca.df = seurat.obj@dr$pca@cell.embeddings
+pca.df = cbind(pca.df, data.frame(Ident = as.vector(seurat.obj@ident)))
+pca.fname = file.path(output_folder_material, 'pca.csv')
+write.csv(pca.df, pca.fname)
+
+print("Performing clustering ... ")
+command = sprintf("%s ./clustering.py %s %s", python.addr, output_folder, n_clusters)
+system(command, wait = T)
+agg.clusters = read.csv(file.path(output_folder_material, 'agglomerative_clustering.csv'))
+colnames(agg.clusters) = c('CellNames', 'ClusterIdx')
+gau.clusters = read.csv(file.path(output_folder_material, 'gaussian_mixture.csv'))
+colnames(gau.clusters) = c('CellNames', 'ClusterIdx')
+metrics_file = file.path(output_folder_material, 'agreement_measures.txt')
+metrics_file = file(metrics_file)
+metrics = readLines(metrics_file)
+close(metrics_file)
+metrics = round(as.numeric(metrics), digits=2)
+metrics = matrix(metrics, nrow=2)
+colnames(metrics) = c('Rand Index', 'Adj. Mutual Information')
+rownames(metrics) = c('Louvain vs Agglomerative', 'Louvain vs Gaussian Mixture')
+metrics = as.table(metrics)
+library(gridExtra)
+library(grid)
+grid.fname = file.path(output_folder, 'metrics.pdf')
+pdf(grid.fname, width = 6, height = 1.5)
+grid.table(metrics)
+dev.off()
+
+seurat.obj@meta.data$gmm = gau.clusters$ClusterIdx
+seurat.obj@meta.data$agg = agg.clusters$ClusterIdx
+
+
+if (!is.na(type.to.colour)){
+  type.to.colour = read.csv(file.path("../../resources", type.to.colour))
+  filter.key   =  type.to.colour$CellTypes %in% as.vector(unique(seurat.obj@ident))
+  cell.labels  = as.vector(type.to.colour$CellTypes[filter.key])
+  cell.colours = as.vector(type.to.colour$Colours[filter.key])
+}else{
+  cell.labels  = sort(as.vector(unique(seurat.obj@ident)))
+  cell.colours = sample(colorRampPalette(brewer.pal(12, "Paired"))(length(cell.labels)))
+}
+
+louvain_idx = as.vector(seurat.obj@ident)
+gmm_idx = as.vector(seurat.obj@meta.data$gmm)
+agg_idx = as.vector(seurat.obj@meta.data$agg)
+umap_data = as.data.frame(seurat.obj@dr$umap@cell.embeddings)
+
+plot.umap.louvain = dr.plot(point.labels=louvain_idx, dr1=umap_data$umap1, dr2=umap_data$umap2, dr1.name="UMAP-x", dr2.name="UMAP-y", no.legend=T, plt.lb.sz=7, txt.lb.size=4, use.cols=cell.colours, use.labels=cell.labels)
+plot.umap.louvain = plot.umap.louvain + ggtitle('Cell types from\nLouvain clustering')
+gmm_types = as.vector(unique(gmm_idx))
+gmm_types = cell.labels[cell.labels %in% gmm_types]
+gmm_colours = mapvalues(x=gmm_types, from=cell.labels, to=cell.colours)
+gmm_index   = seq(1, length(cell.labels))[cell.labels %in% gmm_types]
+plot.umap.gm = dr.plot(point.labels=gmm_idx, dr1=umap_data$umap1, dr2=umap_data$umap2, dr1.name="UMAP-x", dr2.name="UMAP-y", no.legend=T, plt.lb.sz=7, txt.lb.size=4, use.cols=gmm_colours, use.labels=gmm_types, index.map=gmm_index)
+plot.umap.gm = plot.umap.gm + ggtitle('Cell types from\nGaussian Mixture clustering')
+agg_types = as.vector(unique(agg_idx))
+agg_types = cell.labels[cell.labels %in% agg_types]
+agg_colours = mapvalues(x=agg_types, from=cell.labels, to=cell.colours)
+agg_index   = seq(1, length(cell.labels))[cell.labels %in% agg_types]
+plot.umap.agg = dr.plot(point.labels=agg_idx, dr1=umap_data$umap1, dr2=umap_data$umap2, dr1.name="UMAP-x", dr2.name="UMAP-y", no.legend=T, plt.lb.sz=7, txt.lb.size=4, use.cols=agg_colours, use.labels=agg_types, index.map=agg_index)
+plot.umap.agg = plot.umap.agg + ggtitle('Cell types from\nAgglomerative clustering')
+
+dm.fname = file.path(output_folder, 'clustering_comparisons.pdf')
+pdf(dm.fname, width = 18, height = 5)
+plot_grid(plot.umap.louvain, plot.umap.gm, plot.umap.agg, nrow=1)
+dev.off()
+
+n.cols = min(2, length(cell.labels))
+n.rows = ceiling(length(cell.labels) / n.cols)
+plot.legend = plot.indexed.legend(label.vector=cell.labels, color.vector=cell.colours, ncols=n.cols, left.limit=.2, symbol.size=10, text.size=6, padH=.6, padV=.6)
+legend.file.name = file.path(output_folder, 'legend.pdf')
+pdf(legend.file.name, width = 1.5 + .15 * n.cols * max(unlist(lapply(cell.labels, nchar))), height = .5 + n.rows * .35)
+print(plot.legend)
+dev.off()
+
+plot.umap.louvain = dr.plot(point.labels=louvain_idx, dr1=umap_data$umap1, dr2=umap_data$umap2, dr1.name="UMAP-x", dr2.name="UMAP-y", no.legend=T, plt.lb.sz=7, txt.lb.size=4, use.cols=cell.colours, use.labels=cell.labels, annotate.plot=F)
+plot.umap.louvain = plot.umap.louvain + ggtitle('Cell types from\nLouvain clustering')
+gmm_types = as.vector(unique(gmm_idx))
+gmm_types = cell.labels[cell.labels %in% gmm_types]
+gmm_colours = mapvalues(x=gmm_types, from=cell.labels, to=cell.colours)
+gmm_index   = seq(1, length(cell.labels))[cell.labels %in% gmm_types]
+plot.umap.gm = dr.plot(point.labels=gmm_idx, dr1=umap_data$umap1, dr2=umap_data$umap2, dr1.name="UMAP-x", dr2.name="UMAP-y", no.legend=T, plt.lb.sz=7, txt.lb.size=4, use.cols=gmm_colours, use.labels=gmm_types, index.map=gmm_index, annotate.plot=F)
+plot.umap.gm = plot.umap.gm + ggtitle('Cell types from\nGaussian Mixture clustering')
+agg_types = as.vector(unique(agg_idx))
+agg_types = cell.labels[cell.labels %in% agg_types]
+agg_colours = mapvalues(x=agg_types, from=cell.labels, to=cell.colours)
+agg_index   = seq(1, length(cell.labels))[cell.labels %in% agg_types]
+plot.umap.agg = dr.plot(point.labels=agg_idx, dr1=umap_data$umap1, dr2=umap_data$umap2, dr1.name="UMAP-x", dr2.name="UMAP-y", no.legend=T, plt.lb.sz=7, txt.lb.size=4, use.cols=agg_colours, use.labels=agg_types, index.map=agg_index, annotate.plot=F)
+plot.umap.agg = plot.umap.agg + ggtitle('Cell types from\nAgglomerative clustering')
+
+dm.fname = file.path(output_folder, 'clustering_comparisons_no_dots.pdf')
+pdf(dm.fname, width = 18, height = 5)
+plot_grid(plot.umap.louvain, plot.umap.gm, plot.umap.agg, nrow=1)
+dev.off()
+
+unlink(output_folder_material, recursive=T, force=T)
+
+print("Ended beautifully ... ")

pipelines/18_clustering_comparison/clustering_comparison.sh

@@ -0,0 +1,16 @@
+#!/bin/bash
+
+#$ -cwd
+#$ -N clustering_comparison
+#$ -V
+#$ -l h_rt=23:59:59
+#$ -l h_vmem=600G
+
+if [ "$#" -ne 1 ]; then
+    echo "Illegal number of parameters"
+    exit 1
+fi
+
+Rscript clustering_comparison.R $1
+
+echo "End on `date`"