-
Notifications
You must be signed in to change notification settings - Fork 3
/
Clustering_Baseline
114 lines (81 loc) · 4.75 KB
/
Clustering_Baseline
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
## Run baseline clustering and benchmarking:
#Start multiple cluster analysis
#SPATA_to_Seurat
seuratObj <- SPATA::compileSeuratObject(obj,
SCTransform = T,
NormalizeData = T,
FindVariableFeatures = T,
ScaleData = T,
RunPCA = T,
FindNeighbors = T,
FindClusters = T,
RunTSNE = T,
RunUMAP = T)
#SNN-Analysis:
cluster_summary=data.frame(SNN=length(unique([email protected]$seurat_clusters)) )
#Run Graphed based cluster on normal PCA
ggsave("cluster_SNN_TSNE.png",DimPlot(seuratObj, group.by="seurat_clusters", pt.size=2))
#Run Graphed based cluster on normal GLM-PCA
library(glmpca)
library(scry)
library(Seurat)
library(SeuratWrappers)
m <- GetAssayData(seuratObj, slot = "counts", assay = "RNA")
devs <- scry::devianceFeatureSelection(m)
dev_ranked_genes <- rownames(seuratObj)[order(devs, decreasing = TRUE)]
topdev <- head(dev_ranked_genes, 2000)
ndims <- 10
RunGLMPCA2 <- function (object, L = 5, assay = NULL, features = NULL, reduction.name = "glmpca", reduction.key = "GLMPC_", verbose = TRUE, ...) {
#CheckPackage(package = "glmpca", repository = "CRAN")
if (!inherits(x = object, what = "Seurat")) {
stop("'object' must be a Seurat object", call. = FALSE)
}
assay <- assay %||% DefaultAssay(object = object)
DefaultAssay(object = object) <- assay
features <- features %||% VariableFeatures(object)
data <- GetAssayData(object = object, slot = "counts")
features <- intersect(x = features, y = rownames(x = data))
if (length(x = features) == 0) {
stop("Please specify a subset of features for GLM-PCA")
}
data <- data[features, ]
glmpca_results <- glmpca:::glmpca(Y = data, L = L, ...)
glmpca_dimnames <- paste0(reduction.key, 1:L)
colnames(x = glmpca_results$factors) <- glmpca_dimnames
colnames(x = glmpca_results$loadings) <- glmpca_dimnames
factors_l2_norm <- apply(X = glmpca_results$factors, MARGIN = 2,
FUN = function(x) {
sqrt(x = crossprod(x = x))
})
object[[reduction.name]] <- CreateDimReducObject(embeddings = as.matrix(x = glmpca_results$factors),
key = reduction.key, loadings = as.matrix(x = glmpca_results$loadings),
stdev = factors_l2_norm, assay = assay, global = TRUE,
misc = list(glmpca_results))
object <- LogSeuratCommand(object = object)
return(object)
}
seuratObj <- RunGLMPCA2(seuratObj, features = topdev, L = ndims)
seuratObj <- FindNeighbors(seuratObj, reduction = 'glmpca', dims = 1:ndims, verbose = FALSE)
seuratObj <- FindClusters(seuratObj, verbose = FALSE)
seuratObj <- RunTSNE(seuratObj, reduction = 'glmpca', dims = 1:ndims, verbose = FALSE)
ggsave("cluster_GLM_PCA_SNN_TSNE.png",DimPlot(seuratObj, group.by="seurat_clusters", pt.size=2))
cluster_summary$GLM_PCA <- length(unique([email protected]$seurat_clusters))
#PAM Cluster
pam1 <- function(x,k) list(cluster = cluster::pam(x,k, cluster.only=TRUE))
gap_stat_kmeans <- cluster::clusGap(seuratObj@reductions$pca[,1:5],
FUN = kmeans,
nstart = 25,
K.max = 10,
B = 2)
gap_stat_pam <- cluster::clusGap(seuratObj@reductions$pca[,1:5],
FUN = pam1,
K.max = 10,
B = 2)
cluster_summary$kmeans <- which.max(gap_stat_kmeans$Tab %>% as.data.frame() %>% pull(gap) )
cluster_summary$pam <- which.max(gap_stat_pam$Tab %>% as.data.frame() %>% pull(gap) )
cluster_summary <-
cluster_summary %>% t() %>% as.data.frame() %>% rownames_to_column("methods")
ggsave("Cluster_plot.pdf", device="pdf",
ggplot(data=cluster_summary, aes(x=methods, y=V1))+geom_bar(stat="identity", color="grey", alpha=0.5)+theme_classic()
)
ggsave("cluster.png",SPATA::plotSurface(obj, color_to = "seurat_clusters", pt_size=4))