-
Notifications
You must be signed in to change notification settings - Fork 0
/
down_and_perm_2.R
205 lines (183 loc) · 8.61 KB
/
down_and_perm_2.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
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
library("dplyr")
library("Matrix")
library("Seurat")
library("stringr")
library("ggplot2")
geneCap <- function(gene, gene_names) {
# Gene the gene name in the right format
gene_lower <- tolower(gene)
gene_upper <- toupper(gene)
gene_title <- str_to_title(gene)
error <- FALSE
if (gene_lower %in% gene_names) {
gene <- gene_lower
} else if (gene_upper %in% gene_names) {
gene <- gene_upper
} else if (gene_title %in% gene_names) {
gene <- gene_title
} else {
error <- TRUE
}
return(c(gene, error))
}
validGenes <- function(genes, gene_names) {
valid_genes <- c()
for (gene in genes) {
result <- geneCap(gene, gene_names)
gene <- result[1]
error <- as.logical(result[2])
if (! error) {
valid_genes <- c(valid_genes, gene)
}
} # end gene for
return(valid_genes)
} # end validGenes function
downsample <- function(combined, marker_genes, run) {
set.seed(run)
min_trans <- min(combined$nCount_RNA)
gene_names <- rownames(combined@assays$RNA@counts)
# new_matrix <- matrix(, nrow = nrow(combined@assays$RNA@counts), ncol = ncol(combined@assays$RNA@counts), dimnames = list(gene_names, colnames(combined@assays$RNA@counts)))
# new_new_matrix <- matrix(, nrow=nrow(combined@assays$RNA@counts))
marker_matrix <- matrix(, nrow=length(marker_genes), ncol = ncol(combined@assays$RNA@counts), dimnames = list(marker_genes, colnames(combined@assays$RNA@counts)))
i <- 0
for (cell in colnames(combined@assays$RNA@counts)) {
# i <- i + 1
# if (i%%500 == 1) {
# print(cell)
# }
# start.time <- Sys.time()
trans_names <- rep(gene_names, combined@assays$RNA@counts[,cell])
ran_trans_names <- sample(trans_names, min_trans)
ran_trans_names <- ran_trans_names[which(ran_trans_names %in% marker_genes)]
ran_df <- as.data.frame(table(ran_trans_names))
zero_gene_names <- marker_genes[which(! marker_genes %in% ran_trans_names)]
zero_df <- setNames(data.frame(zero_gene_names <- zero_gene_names, Freq <- rep(0, length(zero_gene_names))), c("ran_trans_names", "Freq"))
ran_df <- rbind(ran_df, zero_df)
rownames(ran_df) <- ran_df$ran_trans_names
ran_df <- ran_df[marker_genes,2]
# new_matrix[,cell] <-as.matrix(ran_df)
marker_matrix[,cell] <- as.matrix(ran_df)
# new_new_matrix <- cbind(new_new_matrix, as.matrix(ran_df))
# end.time <- Sys.time()
# time.taken <- end.time - start.time
# print(time.taken)
}
return(marker_matrix)
}
pickNewCells <- function(combined, num_clusters, num_cells) {
new_cells <- c()
for (i in 1:num_cells) {
ran_cluster <- sample(0:num_clusters, 1)
this_cells <- names(combined$seurat_clusters[which(combined$seurat_clusters == ran_cluster)])
new_cells <- c(new_cells, sample(this_cells,1))
}
return(new_cells)
}
shuffleClusters <- function(combined) {
# The selection process for a new cluster should be as follows:
# 1. Pick a random cluster 0-40
# 2. Pick a random cell from that cluster to be a part of the new cluster
# This means that the new data set would likely have duplicate cells
new_cells <- lapply(0:num_clusters, function(x) c())
num_clusters <- as.numeric(tail(levels([email protected]$seurat_clusters), n=1))
for (i in 0:num_clusters) {
num_cells <- length(combined$seurat_clusters[which(combined$seurat_clusters == i)])
new_cells[[i+1]] <- pickNewCells(combined, num_clusters, num_cells)
}
return(new_cells)
}
## END FUNCTIONS ##
# rna_path <- "C:/Users/miles/Downloads/brain/"
rna_path <- "~/scratch/brain/"
# combined <- readRDS(paste(rna_path, "/brain_scripts/brain_shiny/data/combined.rds", sep = ""))
combined <- readRDS(paste(rna_path, "/data/B1C1C2MZ_combined_031020.rds", sep=""))
marker_path <- paste(rna_path, "data/markers/", sep="")
marker_files <- dir(marker_path, pattern =paste("*.txt", sep=""))
markers <- data.frame(gene <- c(), bio <- c())
for (i in 1:length(marker_files)) {
file <- read.table(paste(marker_path, marker_files[i], sep=""), header=FALSE, sep="\t", stringsAsFactors=FALSE)
markers <- rbind(markers, file[,1:2])
}
colnames(markers) <- c("gene", "bio")
bio <- "ROCK_SAND"
markers <- markers[which(markers$bio == bio),]
gene_names <- rownames(combined@assays$RNA)
marker_genes <- markers$gene
valid_genes <- marker_genes
num_clusters <- as.numeric(tail(levels([email protected]$seurat_clusters), n=1))
down_avg_avg_gene <- rep(0, num_clusters+1)
run_num <- 50
# No Perm, Bootstrap
down_list <- lapply(0:num_clusters, function(x) c())
for (run in 1:run_num) {
cat(paste("no_perm", run, "\n"))
mat <- downsample(combined, marker_genes, run)
cells_per_cluster <- c()
genes_per_cluster <- c()
for (i in 0:num_clusters) {
this_cells <- WhichCells(combined, idents = i)
this_genes <- length(which(as.vector(mat[,this_cells]) != 0))
# genes_per_cluster <- c(genes_per_cluster, length(which(as.vector(combined@assays$RNA@counts[ran_markers,this_cells]) != 0))) # genes
genes_per_cluster <- c(genes_per_cluster, this_genes) # genes
cells_per_cluster <- c(cells_per_cluster, length(this_cells))
down_list[[i+1]] <- c(down_list[[i+1]], this_genes/ length(this_cells))
}
avg_gene_per_cell_per_cluster <- genes_per_cluster/cells_per_cluster
down_avg_avg_gene <- down_avg_avg_gene + avg_gene_per_cell_per_cluster
}
down_avg_avg_gene <- down_avg_avg_gene / run_num
print(down_avg_avg_gene)
# Perm, Bootstrap
backup_ids <- [email protected]$seurat_clusters
perm_down_avg_gene <- lapply(0:num_clusters, function(x) c())
for (run in (run_num+1):(run_num+run_num)) {
cat(paste("perm", run, "\n"))
set.seed(run)
mat <- downsample(combined, marker_genes, run)
new_cells <- shuffleClusters(combined)
num_clusters <- as.numeric(tail(levels([email protected]$seurat_clusters), n=1))
gene_names <- rownames(combined@assays$RNA)
cells_per_cluster <- c()
genes_per_cluster <- c()
for (i in 0:num_clusters) {
this_cells <- new_cells[[i+1]]
this_genes <- length(which(as.vector(mat[valid_genes,this_cells]) != 0))
genes_per_cluster <- c(genes_per_cluster, this_genes) # genes
cells_per_cluster <- c(cells_per_cluster, length(this_cells))
perm_down_avg_gene[[i+1]] <- c(perm_down_avg_gene[[i+1]], this_genes/length(this_cells))
}
avg_gene_per_cell_per_cluster <- genes_per_cluster/cells_per_cluster
# perm_down_avg_gene <- c(perm_down_avg_gene, avg_gene_per_cell_per_cluster)
}
# Compare empirical data to 97.5th percentile of the permutated data on a PER CLUSTER basis
sig_clusters <- c()
alpha <- 0.01
bonferroni_alpha <- (alpha/num_clusters)/length(gene_names)
# upper_tail <- bonferroni_alpha/2
upper_tail <- alpha/2
for (i in 0:num_clusters) {
sig <- quantile(perm_down_avg_gene[[i+1]], c(1-upper_tail))
print(paste("Max value for perm cluster ", i, ":", max(perm_down_avg_gene[[i+1]])))
print(paste("Mean Real value for cluster", i, ":", down_avg_avg_gene[i+1]))
print(paste("Min Real Value for cluster", i, ":", min(down_list[[i+1]])))
bot_25 <- quantile(down_list[[i+1]], c(0.25))
print(paste("Bot 25 Value for cluster", i, ":", bot_25))
if ( min(down_list[[i+1]]) > max(perm_down_avg_gene[[i+1]]) ) {
sig_clusters <- c(sig_clusters, i)
}
}
# sig <- quantile(perm_down_avg_gene, c(.975))
# print(sig)
# sig_clusters <- which(down_avg_avg_gene > sig)-1
df <- t(as.data.frame(lapply(1:length(down_list), function(x) t(cbind(down_list[[x]], rep(x, length(num_clusters)+1), rep("Down", length(num_clusters)+1))))))
df <- rbind(df, t(as.data.frame(lapply(1:length(perm_down_avg_gene), function(x) t(cbind(perm_down_avg_gene[[x]], rep(x, length(num_clusters)+1), rep("Down+Perm", length(num_clusters)+1)))))))
colnames(df) <- c("avg_gene_per_cell_per_cluster", "cluster", "cond")
df <- as.data.frame(df)
df$avg_gene_per_cell_per_cluster <- as.numeric(as.vector(df$avg_gene_per_cell_per_cluster))
df$cluster <- factor(df$cluster, levels = 1:(num_clusters+1))
png(paste0(rna_path, "/results/down_and_perm_2_", bio, ".png"), width = 1800, height = 1000, res = 150)
p <- ggplot(df, aes(x = cluster, y = avg_gene_per_cell_per_cluster, fill = cond)) + geom_boxplot(alpha = 0.6) + geom_jitter(shape=16, position=position_jitterdodge(), alpha = 0.3, aes(colour = cond)) + scale_colour_manual(values=c("#999999", "#56B4E9", "#3ac9bb")) + scale_fill_manual(values=c("#999999", "#3ac9bb", "#56B4E9")) + ggtitle(paste(bio, "- Average Genes per Cell per Cluster"))
print(p)
dev.off()
print(sig_clusters)
write.csv(sig_clusters, file = paste(rna_path, "/results/down_and_perm_2_", bio, ".csv", sep=""), row.names = FALSE)