-
Notifications
You must be signed in to change notification settings - Fork 2
/
Subcluster_harmony.R
133 lines (106 loc) · 6.01 KB
/
Subcluster_harmony.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
require("Seurat")
liver.integrated <- readRDS("integration_harmony_plus_analysis.rds")
set.seed(7742)
# Dimensionality Reduction
require("sctransform")
#liver.integrated <- ScaleData(liver.integrated);
#liver.integrated <- RunPCA(liver.integrated, features = VariableFeatures(object = liver.integrated))
ElbowPlot(liver.integrated)
npcs <- 20
# Cluster with many different parameters
res <- seq(from=0.3, to=2, by=0.2)
nkNN <- seq(from=30, to=90, 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)
name <- paste("knn_",nkNN_param,"_res_", res_param, sep="");
liver.integrated@meta.data[[name]] <- liver.integrated@meta.data$seurat_clusters;
}}
# Compare all these clusterings
require(igraph)
require(gplots)
clust_table <- liver.integrated@meta.data[, grepl("^knn_", names(liver.integrated@meta.data))]
clust_table <- as.matrix(apply(clust_table,2,as.numeric))
require("proxy")
clust_dists <- proxy::dist(clust_table, method=function(x,y){igraph::compare(x,y,method="vi")}, by_rows=FALSE)
clust_similr1 <- proxy::simil(clust_table, method=function(x,y){igraph::compare(x,y,method="nmi")}, by_rows=FALSE)
clust_similr2 <- proxy::simil(clust_table, method=function(x,y){igraph::compare(x,y,method="adjusted.rand")}, by_rows=FALSE)
# Find robust exemplar clustering(s)
require("apcluster")
require("gplots")
set.seed(18371)
res1 <- apcluster(-1*as.matrix(clust_dists), p=-2.5)
#res2 <- apcluster(as.matrix(clust_similr1), p=-2)
#res3 <- apcluster(as.matrix(clust_similr1), p=-2)
#valid_clusterings <- res1@exemplars[which(res1@exemplars %in% res2@exemplars & res1@exemplars %in% res3@exemplars)]
coarse_lvl <- "knn_70_res_0.3" # basic_integration_analysis.rds = knn_70_res_0.3
fine_lvl <- "knn_70_res_0.9" # basic_integration_analysis.rds = knn_90_res_1.5
#manually select which exemplar to use
liver.integrated@meta.data$Coarse_clusters <- liver.integrated@meta.data[[coarse_lvl]]
liver.integrated@meta.data$Fine_clusters <- liver.integrated@meta.data[[fine_lvl]]
apcluster::heatmap(res1, -1*as.matrix(clust_dists))
png("compare_harmony_clusterings_heatmap.png", width=6, height=6, units="in", res=300)
lab <- matrix("", ncol=ncol(clust_table), nrow=ncol(clust_table))
lab[colnames(clust_table)==fine_lvl, colnames(clust_table)==fine_lvl] <- "F"
lab[colnames(clust_table)==coarse_lvl, colnames(clust_table)==coarse_lvl] <- "C"
heatmap.2(as.matrix(clust_dists), trace="none", distfun=function(x){return(as.dist(clust_dists))}, cellnote=lab)
dev.off()
# Visualize the Chosen clusterings
nkNN <- 70
res <- 0.9
liver.integrated <- RunTSNE(liver.integrated, reduction="harmony", dims = 1:npcs)
liver.integrated <- RunUMAP(liver.integrated, reduction="harmony", dims = 1:npcs, parallel=FALSE, n.neighbour=nkNN)
png("Highres_harmony_coarse_integrated_umap.png", width=6, height=6, units="in", res=50)
DimPlot(liver.integrated, reduction = "umap", group.by="Coarse_clusters")
dev.off()
png("Highres_harmony_coarse_integrated_tsne.png", width=6, height=6, units="in", res=50)
DimPlot(liver.integrated, reduction = "tsne", group.by="Coarse_clusters")
dev.off()
png("Highres_harmony_fine_integrated_umap.png", width=6, height=6, units="in", res=50)
DimPlot(liver.integrated, reduction = "umap", group.by="Fine_clusters")
dev.off()
png("Highres_harmony_fine_integrated_tsne.png", width=6, height=6, units="in", res=50)
DimPlot(liver.integrated, reduction = "tsne", group.by="Fine_clusters")
dev.off()
png("Donor_harmony_integrated_umap.png", width=6, height=6, units="in", res=50)
DimPlot(liver.integrated, reduction = "umap", group.by="orig.ident")
dev.off()
png("Donor_harmony_integrated_tsne.png", width=6, height=6, units="in", res=50)
DimPlot(liver.integrated, reduction = "tsne", group.by="orig.ident")
dev.off()
png("Coarse_harmony_clusters_by_donor.png", width=6, height=6, units="in", res=50)
barplot(table(liver.integrated@meta.data$orig.ident, liver.integrated@meta.data$Coarse_clusters), col=rainbow(20))
dev.off()
png("Fine_harmony_clusters_by_donor.png", width=6, height=6, units="in", res=50)
barplot(table(liver.integrated@meta.data$orig.ident, liver.integrated@meta.data$Fine_clusters), col=rainbow(20))
dev.off()
# Auto-annotation
source("Setup_autoannotation.R")
all_anno <- readRDS("All20_automatedannotation.rds");
liver.integrated@meta.data$scmap_anno <- rep("unknown", ncol(liver.integrated));
liver.integrated@meta.data$scmap_anno2 <- rep("unknown", ncol(liver.integrated));
for (donor in unique(liver.integrated@meta.data$orig.ident)) {
cell_ids <- liver.integrated@meta.data$cell_barcode[liver.integrated@meta.data$donor == donor]
anno <- all_anno[[donor]];
anno <- anno[anno$cell_barcode %in% cell_ids,]
anno <- anno[match(cell_ids, anno$cell_barcode),]
liver.integrated@meta.data$scmap_anno[liver.integrated@meta.data$orig.ident == donor] <- as.character(anno$scmap_cluster_anno$lm1)
liver.integrated@meta.data$scmap_anno2[liver.integrated@meta.data$orig.ident == donor] <- as.character(anno$scmap_cell_anno)
}
png("AutoAnno_harmony_integrated_umap.png", width=8, height=6, units="in", res=50)
DimPlot(liver.integrated, reduction = "umap", group.by="scmap_anno")
dev.off()
png("AutoAnno_harmony_integrated_tsne.png", width=8, height=6, units="in", res=50)
DimPlot(liver.integrated, reduction = "tsne", group.by="scmap_anno")
dev.off()
png("AutoAnno2_harmony_integrated_umap.png", width=8, height=6, units="in", res=50)
DimPlot(liver.integrated, reduction = "umap", group.by="scmap_anno2")
dev.off()
png("AutoAnno2_harmony_integrated_tsne.png", width=8, height=6, units="in", res=50)
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")