-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathMap2.2_New_Select_Clustering.R
174 lines (140 loc) · 7.84 KB
/
Map2.2_New_Select_Clustering.R
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
require("Seurat")
args <- commandArgs(trailingOnly=TRUE)
#proj_name="Map2.2_merged_integrated"
proj_name=args[1]
#liver.integrated <- readRDS("harmony_integrated.rds")
#liver.integrated <- readRDS("integration_v2.rds")
#liver.integrated <- readRDS("Merged_integrated_with_metadata.rds")
#liver.integrated <- readRDS("Map2.2_merged_integrated.rds")
liver.integrated <- readRDS(args[2])
set.seed(7742)
# Dimensionality Reduction
#require("sctransform")
#liver.integrated <- ScaleData(liver.integrated);
#liver.integrated <- RunPCA(liver.integrated, features = VariableFeatures(object = liver.integrated))
npcs <- 15 # was 20 - 29/04/2020
require("ggplot2")
png(paste(proj_name,"pca_elbow.png", sep="_"), width=4, height=4, units="in", res=100)
a <- ElbowPlot(liver.integrated)+geom_vline(aes(xintercept=npcs), linetype="dotted", color="grey35", size=2)
plot(a)
dev.off()
# Cluster with many different parameters
res <- seq(from=0.3, to=2, by=0.3)
nkNN <- seq(from=40, to=80, by=10)
for(res_param in res) {
for(nkNN_param in nkNN){
liver.integrated <- FindNeighbors(liver.integrated, reduction="harmony", dims = 1:npcs, k.param=nkNN_param)
#liver.integrated <- FindClusters(liver.integrated, reduction="harmony", resolution = res_param, k.param=nkNN_param)
liver.integrated <- FindClusters(liver.integrated, reduction="harmony", resolution = res_param, group.singletons=FALSE, method="igraph") # for UHN Seurat installation
name <- paste("knn_",nkNN_param,"_res_", res_param, sep="");
[email protected][[name]] <- [email protected]$seurat_clusters;
}}
saveRDS(liver.integrated, paste(proj_name, "harmony_plus_clusters.rds", sep="_"))
# Compare all these clusterings
require(igraph)
require(gplots)
clust_table <- [email protected][, grepl("^knn_", colnames([email protected]))]
clust_table <- as.matrix(apply(clust_table,2,as.numeric))
require("proxy")
clust_dists <- proxy::dist(clust_table, method=function(x,y){igraph::compare(x,y,method="vi")}, by_rows=FALSE)
#clust_similr1 <- proxy::simil(clust_table, method=function(x,y){igraph::compare(x,y,method="nmi")}, by_rows=FALSE)
#clust_similr2 <- proxy::simil(clust_table, method=function(x,y){igraph::compare(x,y,method="adjusted.rand")}, by_rows=FALSE)
# Find robust exemplar clustering(s)
require("apcluster")
require("gplots")
set.seed(18371)
res1 <- apcluster(-1*as.matrix(clust_dists), p=-2.5) # SoupX Map = -2.2, Empty Map = -2.5
#res2 <- apcluster(as.matrix(clust_similr1), p=-2)
#res3 <- apcluster(as.matrix(clust_similr1), p=-2)
#valid_clusterings <- res1@exemplars[which(res1@exemplars %in% res2@exemplars & res1@exemplars %in% res3@exemplars)]
# Cluster-Cell-type Purity
type_purity <- function(clusters, annotations) {
tmp <- table(clusters, annotations)
shan <- median(vegan::diversity(tmp, index="shannon", MARGIN=1))
simp <- median(vegan::diversity(tmp, index="simpson", MARGIN=1))
return(c(shan, simp))
}
score <- apply(clust_table, 2, type_purity, [email protected]$marker_labs)
#plot(score, xlab="clustering" ylab="type_purity")
score2 <- colMeans(score)
exemplars <- c();
for (c in res1@clusters) {
clusterings <- names(c)
scores <- score2[clusterings]
exemplars <- c(exemplars, scores[which(scores == min(scores))])
}
liver.integrated@misc$exemplars <- exemplars;
core_lvl <- names(exemplars)[1]
coarse_lvl <- names(exemplars)[2]
fine_lvl <- names(exemplars)[3]
#coarse_lvl <- "knn_70_res_0.3" # basic_integration_analysis.rds = knn_70_res_0.3
#fine_lvl <- "knn_70_res_0.9" # basic_integration_analysis.rds = knn_90_res_1.5
#manually select which exemplar to use
[email protected]$Core_clusters <- [email protected][[core_lvl]]
[email protected]$Coarse_clusters <- [email protected][[coarse_lvl]]
[email protected]$Fine_clusters <- [email protected][[fine_lvl]]
# Start with the Core clusters - Do demographic DE on them + Hepatocyte tranjectory
# Subcluster the Core clusters - including all the genes.
#apcluster::heatmap(res1, -1*as.matrix(clust_dists))
png(paste(proj_name,"compare_clusterings_heatmap.png",sep="_"), width=6, height=6, units="in", res=300)
lab <- matrix("", ncol=ncol(clust_table), nrow=ncol(clust_table))
lab[colnames(clust_table)==fine_lvl, colnames(clust_table)==fine_lvl] <- "3"
lab[colnames(clust_table)==coarse_lvl, colnames(clust_table)==coarse_lvl] <- "2"
lab[colnames(clust_table)==core_lvl, colnames(clust_table)==core_lvl] <- "1"
heatmap.2(as.matrix(clust_dists), trace="none", distfun=function(x){return(as.dist(clust_dists))}, cellnote=lab)
dev.off()
# Visualize the Chosen clusterings
#nkNN <- 70
png(paste(proj_name,"core_umap.png", sep="_"), width=6, height=6, units="in", res=150)
DimPlot(liver.integrated, reduction = "umap", group.by="Core_clusters")
dev.off()
#liver.integrated <- RunTSNE(liver.integrated, reduction="harmony", dims = 1:npcs)
#liver.integrated <- RunUMAP(liver.integrated, reduction="harmony", dims = 1:npcs, parallel=FALSE, n.neighbour=nkNN)
png(paste(proj_name,"coarse_umap.png", sep="_"), width=6, height=6, units="in", res=150)
DimPlot(liver.integrated, reduction = "umap", group.by="Coarse_clusters")
dev.off()
#png(paste(proj_name,"coarse_tsne.png",sep="_"), width=6, height=6, units="in", res=150)
#DimPlot(liver.integrated, reduction = "tsne", group.by="Coarse_clusters")
#dev.off()
png(paste(proj_name,"fine_umap.png", sep="_"), width=6, height=6, units="in", res=150)
DimPlot(liver.integrated, reduction = "umap", group.by="Fine_clusters")
dev.off()
#png(paste(proj_name,"fine_tsne.png", sep="_"), width=6, height=6, units="in", res=150)
#DimPlot(liver.integrated, reduction = "tsne", group.by="Fine_clusters")
#dev.off()
saveRDS(liver.integrated, paste(proj_name, "harmony_plus_analysis.rds", sep="_"))
#png("Coarse_harmony_clusters_by_donor.png", width=6, height=6, units="in", res=150)
#barplot(table([email protected]$orig.ident, [email protected]$Coarse_clusters), col=rainbow(20))
#dev.off()
#png("Fine_harmony_clusters_by_donor.png", width=6, height=6, units="in", res=150)
#barplot(table([email protected]$orig.ident, [email protected]$Fine_clusters), col=rainbow(20))
#dev.off()
# Auto-annotation
#source("Setup_autoannotation.R")
#all_anno <- readRDS("All20_automatedannotation.rds");
#[email protected]$scmap_anno <- rep("unknown", ncol(liver.integrated));
#[email protected]$scmap_anno2 <- rep("unknown", ncol(liver.integrated));
#for (donor in unique([email protected]$orig.ident)) {
# cell_ids <- [email protected]$cell_barcode[[email protected]$donor == donor]
# anno <- all_anno[[donor]];
# anno <- anno[anno$cell_barcode %in% cell_ids,]
# anno <- anno[match(cell_ids, anno$cell_barcode),]
# [email protected]$scmap_anno[[email protected]$orig.ident == donor] <- as.character(anno$scmap_cluster_anno$lm1)
# [email protected]$scmap_anno2[[email protected]$orig.ident == donor] <- as.character(anno$scmap_cell_anno)
#}
#png("AutoAnno_harmony_integrated_umap.png", width=8, height=6, units="in", res=150)
#DimPlot(liver.integrated, reduction = "umap", group.by="scmap_anno")
#dev.off()
#png("AutoAnno_harmony_integrated_tsne.png", width=8, height=6, units="in", res=150)
#DimPlot(liver.integrated, reduction = "tsne", group.by="scmap_anno")
#dev.off()
#png("AutoAnno2_harmony_integrated_umap.png", width=8, height=6, units="in", res=150)
#DimPlot(liver.integrated, reduction = "umap", group.by="scmap_anno2")
#dev.off()
#png("AutoAnno2_harmony_integrated_tsne.png", width=8, height=6, units="in", res=150)
#DimPlot(liver.integrated, reduction = "tsne", group.by="scmap_anno2")
#dev.off()
#saveRDS(liver.integrated, "integration_harmony_plus_analysis.rds")
# Would this be helpful?
# create a heatmap where: cell = average similarity of this clustering to all other clusterings
# distance between clusterings is measured using igraph::compare(c1, c2, method="vi")