-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathGSEA_ties_simulation_20240909_v2.R
259 lines (225 loc) · 14.9 KB
/
GSEA_ties_simulation_20240909_v2.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
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
# Author: Belinda B. Garana
# can GSEA_ties detect enrichment of synthetic gene set?
### Step 1: Generate gene rank metrics that have a normal distribution, then perturb 1 set of X genes
### Step 2: Run GSEA_ties for each different variant (+/- Y gene rank)
### Step 3: Re-run for 50 replicates
### Step 4: Repeat with different values of X (5, 10, 20, 40)
rm(list=ls(all=TRUE))
if(!require(devtools)){install.packages("dev.tools")}
if(!require(DMEA)){devtools::install_github('BelindaBGarana/DMEA')}
library(DMEA);library(dplyr);library(GSA);library(reshape2);library(data.table);library(ggplot2);library(msigdbr)
setwd("~/OneDrive - PNNL/Documents/GitHub/Chr8/proteomics/analysis/Chr8_quant")
### Step 0: Prep rows, columns, values
plot.data <- read.csv("Proteomics/Global/Differential_expression/Differential_expression_results.csv")
plot.data <- na.omit(plot.data[,c("Gene", "Spearman.est")])
plot.data <- plot.data[plot.data$Spearman.est != 0,]
# shuffle gene names so that no gene set should be enriched unless synthetically
plot.data$Gene <- sample(plot.data$Gene)
Gene <- sample(plot.data$Gene)
gene.names <- Gene
# # model distribution based on real correlation
# hist(plot.data$Spearman.est)
# sd.spearman <- sd(plot.data$Spearman.est) # 0.4421582
# mean.spearman <- mean(plot.data$Spearman.est) # -0.06911027
# norm.dist <- rnorm(nrow(plot.data), mean=mean.spearman, sd = sd.spearman)
# plot(density(plot.data$Spearman.est))
# lines(density(norm.dist), col="red") # close enough
# median number of ties is 90, avg is 101.032287
# create normal distribution of 100 and then add 90 duplicates for total of 9000
Spearman.est <- rnorm(100, mean = 0, sd = 0.5)
Spearman.est <- Spearman.est[abs(Spearman.est) < 1]
Spearman.est <- rep(Spearman.est, length.out = length(Gene))
norm.df <- data.frame(Gene, Spearman.est)
# How far do we want to vary values here?
abs.vary <- 0.25
values.to.vary <- round(seq(from = -abs.vary, to = abs.vary, by = abs.vary/5), digits=2)
synthetic.cell.names <- "Spearman.est"
# Import MOA set information
gmt.info <- msigdbr::msigdbr(category="H") # hallmark gene sets for homo sapiens
gmt <- list("genesets" = list(), "geneset.names" = list(), "geneset.descriptions" = list())
gs <- unique(gmt.info$gs_name)
for (i in 1:length(gs)) {
gmt$genesets[[i]] <- gmt.info[gmt.info$gs_name == gs[i],]$gene_symbol
gmt$geneset.names[[i]] <- gs[i]
gmt$geneset.descriptions[[i]] <- gs[i]
}
path.sim <- paste0("GSEA_ties_sim_results_",Sys.Date())
dir.create(path.sim)
# Change path to where you want files saved
setwd(path.sim)
setwd("valueAdd0.25")
og.all.synth.sim.results <- list()
og.all.sim.results <- list()
og.all.perc.sig <- list()
all.synth.sim.results <- list()
all.sim.results <- list()
all.perc.sig <- list()
synth.gene.sets <- data.frame(N_genes=integer(), synthetic.gene.set=character())
gene.set.sizes <- seq(30, 300, length.out=10)
for(i in 1:(length(gene.set.sizes))){
synthetic.gene.set <- sample(gene.names,gene.set.sizes[i])
synthetic.gene.df <- as.data.frame(synthetic.gene.set)
synthetic.gene.df$N_genes <- gene.set.sizes[i]
synth.gene.sets <- rbind(synth.gene.sets, synthetic.gene.df)
# make new gmt with synthetic gene set
new.gene.sets <- gmt$genesets
new.gene.set.names <- gmt$geneset.names
new.gene.set.descr <- gmt$geneset.descriptions
new.gene.set.names[[length(gmt$genesets)+1]] <- paste0(gene.set.sizes[i],"_random_genes")
new.gene.set.descr[[length(gmt$genesets)+1]] <- paste0(gene.set.sizes[i],"_random_genes")
new.gene.sets[[length(gmt$genesets)+1]] <- synthetic.gene.set
new.gmt <- list(genesets=new.gene.sets, geneset.names=new.gene.set.names, geneset.descriptions=new.gene.set.descr)
og.synth.sim.results <- list()
og.sim.results <- list()
synth.sim.results <- list()
sim.results <- list()
for(N in 1:20){
### Step 3: Generate gene AUC that have a normal distribution for each cell line, then perturb synthetic gene set
og.synthetic.results <- as.data.frame(values.to.vary)
og.synthetic.results[,c("ES","NES","p","q")] <- NA
og.results <- list()
synthetic.results <- as.data.frame(values.to.vary)
synthetic.results[,c("ES","NES","p","q")] <- NA
results <- list()
for(j in 1:length(values.to.vary)){
rank.matrix <- norm.df
row.names(rank.matrix) <- rank.matrix$Gene
# perturb synthetic gene set
curr.vals <- rank.matrix[synthetic.gene.set,synthetic.cell.names]
rank.matrix[synthetic.gene.set,synthetic.cell.names] <- curr.vals + values.to.vary[j]
# run GSEA accounting for ties
DMEA.result <- drugSEA_ties(data = rank.matrix, drug = "Gene", gmt = new.gmt, rank.metric=synthetic.cell.names, plots=FALSE, min.per.set=5)
# i = 1 (5 genes), j=7 (varied val +0.75), N=2: Error in if (temp.NES >= 0) { : missing value where TRUE/FALSE needed
og.DMEA.result <- DMEA.result$result.w.ties
og.synthetic.results[j, c("ES", "NES", "p", "q")] <- og.DMEA.result[
og.DMEA.result$Drug_set==paste0(gene.set.sizes[i],"_random_genes"), c("ES", "NES", "p_value", "FDR_q_value")]
og.results[[as.character(values.to.vary[j])]] <- og.DMEA.result
DMEA.result <- DMEA.result$result
synthetic.results[j, c("ES", "NES", "p", "q")] <- DMEA.result[
DMEA.result$Drug_set==paste0(gene.set.sizes[i],"_random_genes"), c("ES", "NES", "p_value", "FDR_q_value")]
results[[as.character(values.to.vary[j])]] <- DMEA.result
}
og.synth.sim.results[[N]] <- og.synthetic.results
og.sim.results[[N]] <- data.table::rbindlist(og.results, use.names=TRUE, idcol="Value Added", fill=TRUE)
synth.sim.results[[N]] <- synthetic.results
sim.results[[N]] <- data.table::rbindlist(results, use.names=TRUE, idcol="Value Added", fill=TRUE)
}
og.all.synth.sim.results[[as.character(gene.set.sizes[i])]] <- data.table::rbindlist(og.synth.sim.results, use.names=TRUE, idcol="Simulation Number", fill=TRUE)
og.all.sim.results[[as.character(gene.set.sizes[i])]] <- data.table::rbindlist(og.sim.results, use.names=TRUE, idcol="Simulation Number", fill=TRUE)
write.csv(og.all.sim.results[[as.character(gene.set.sizes[i])]],file=paste0(gene.set.sizes[i],"_genes_",N,"_replicates_all_GSEA_ties_results_originalOrder_",Sys.Date(),".csv"))
write.csv(og.all.synth.sim.results[[as.character(gene.set.sizes[i])]],file=paste0(gene.set.sizes[i],"_genes_",N,"_replicates_Synthetic_gene_Set_GSEA_ties_results_originalOrder_",Sys.Date(),".csv"))
og.all.synth.sim.results.df <- og.all.synth.sim.results[[as.character(gene.set.sizes[i])]]
all.synth.sim.results[[as.character(gene.set.sizes[i])]] <- data.table::rbindlist(synth.sim.results, use.names=TRUE, idcol="Simulation Number", fill=TRUE)
all.sim.results[[as.character(gene.set.sizes[i])]] <- data.table::rbindlist(sim.results, use.names=TRUE, idcol="Simulation Number", fill=TRUE)
write.csv(all.sim.results[[as.character(gene.set.sizes[i])]],file=paste0(gene.set.sizes[i],"_genes_",N,"_replicates_all_GSEA_ties_results_",Sys.Date(),".csv"))
write.csv(all.synth.sim.results[[as.character(gene.set.sizes[i])]],file=paste0(gene.set.sizes[i],"_genes_",N,"_replicates_Synthetic_gene_Set_GSEA_ties_results_",Sys.Date(),".csv"))
all.synth.sim.results.df <- all.synth.sim.results[[as.character(gene.set.sizes[i])]]
perc.sig <- as.data.frame(values.to.vary)
perc.sig[,c("N.replicates","avg.NES","sd.NES",
"perc.sig","perc.sig.pos","perc.sig.neg",
"perc.moreSig","perc.moreSig.pos","perc.moreSig.neg")] <- NA
og.perc.sig <- perc.sig
for(k in 1:length(values.to.vary)){
temp.data <- all.synth.sim.results.df[all.synth.sim.results.df$values.to.vary==values.to.vary[k],]
og.temp.data <- og.all.synth.sim.results.df[og.all.synth.sim.results.df$values.to.vary==values.to.vary[k],]
# N and NES
og.perc.sig$N.replicates[k] <- nrow(og.temp.data)
og.perc.sig$avg.NES[k] <- mean(as.numeric(og.temp.data$NES))
og.perc.sig$sd.NES[k] <- sd(as.numeric(og.temp.data$NES))
perc.sig$N.replicates[k] <- nrow(temp.data)
perc.sig$avg.NES[k] <- mean(as.numeric(temp.data$NES))
perc.sig$sd.NES[k] <- sd(as.numeric(temp.data$NES))
# % significant
## og
# p < 0.05 & q < 0.25
og.sig.temp.data <- og.temp.data[og.temp.data$p<0.05 & og.temp.data$q<0.25,]
og.pos.sig.temp.data <- og.sig.temp.data[og.sig.temp.data$NES>0,]
og.neg.sig.temp.data <- og.sig.temp.data[og.sig.temp.data$NES<0,]
og.perc.sig$perc.sig[k] <- 100*nrow(og.sig.temp.data)/nrow(og.temp.data)
og.perc.sig$perc.sig.pos[k] <- 100*nrow(og.pos.sig.temp.data)/nrow(og.temp.data)
og.perc.sig$perc.sig.neg[k] <- 100*nrow(og.neg.sig.temp.data)/nrow(og.temp.data)
# p < 0.05 & q < 0.25
og.sig.temp.data <- og.temp.data[og.temp.data$p<0.05 & og.temp.data$q<0.05,]
og.pos.sig.temp.data <- og.sig.temp.data[og.sig.temp.data$NES>0,]
og.neg.sig.temp.data <- og.sig.temp.data[og.sig.temp.data$NES<0,]
og.perc.sig$perc.moreSig[k] <- 100*nrow(og.sig.temp.data)/nrow(og.temp.data)
og.perc.sig$perc.moreSig.pos[k] <- 100*nrow(og.pos.sig.temp.data)/nrow(og.temp.data)
og.perc.sig$perc.moreSig.neg[k] <- 100*nrow(og.neg.sig.temp.data)/nrow(og.temp.data)
## shuffled ties
# p < 0.05 & q < 0.25
sig.temp.data <- temp.data[temp.data$p<0.05 & temp.data$q<0.25,]
pos.sig.temp.data <- sig.temp.data[sig.temp.data$NES>0,]
neg.sig.temp.data <- sig.temp.data[sig.temp.data$NES<0,]
perc.sig$perc.sig[k] <- 100*nrow(sig.temp.data)/nrow(temp.data)
perc.sig$perc.sig.pos[k] <- 100*nrow(pos.sig.temp.data)/nrow(temp.data)
perc.sig$perc.sig.neg[k] <- 100*nrow(neg.sig.temp.data)/nrow(temp.data)
# p < 0.05 & q < 0.25
sig.temp.data <- temp.data[temp.data$p<0.05 & temp.data$q<0.05,]
pos.sig.temp.data <- sig.temp.data[sig.temp.data$NES>0,]
neg.sig.temp.data <- sig.temp.data[sig.temp.data$NES<0,]
perc.sig$perc.moreSig[k] <- 100*nrow(sig.temp.data)/nrow(temp.data)
perc.sig$perc.moreSig.pos[k] <- 100*nrow(pos.sig.temp.data)/nrow(temp.data)
perc.sig$perc.moreSig.neg[k] <- 100*nrow(neg.sig.temp.data)/nrow(temp.data)
}
write.csv(og.perc.sig,file=paste0(gene.set.sizes[i],"_genes_",N,"_replicates_percent_significant_synthetic_gene_set_enrichments_originalOrder.csv"))
og.all.perc.sig[[as.character(gene.set.sizes[i])]] <- og.perc.sig
write.csv(perc.sig,file=paste0(gene.set.sizes[i],"_genes_",N,"_replicates_percent_significant_synthetic_gene_set_enrichments.csv"))
all.perc.sig[[as.character(gene.set.sizes[i])]] <- perc.sig
}
og.complete.synth.sim.results <- data.table::rbindlist(og.all.synth.sim.results, use.names=TRUE, idcol="Number of genes in Set", fill=TRUE)
og.complete.sim.results <- data.table::rbindlist(og.all.sim.results, use.names=TRUE, idcol="Number of genes in Set", fill=TRUE)
write.csv(og.complete.sim.results,file=paste0(N,"_replicates_all_GSEA_ties_results_simulation_originalOrder_",Sys.Date(),".csv"), row.names=FALSE)
write.csv(og.complete.synth.sim.results,file=paste0(N,"_replicates_synthetic_gene_Set_GSEA_ties_results_originalOrder_",Sys.Date(),".csv"), row.names=FALSE)
og.complete.perc.sig <- data.table::rbindlist(og.all.perc.sig, use.names=TRUE, idcol="Number of genes in Set", fill=TRUE)
write.csv(og.complete.perc.sig,file=paste0(N,"_replicates_percent_significant_synthetic_gene_set_enrichments_originalOrder_",Sys.Date(),".csv"), row.names=FALSE)
complete.synth.sim.results <- data.table::rbindlist(all.synth.sim.results, use.names=TRUE, idcol="Number of genes in Set", fill=TRUE)
complete.sim.results <- data.table::rbindlist(all.sim.results, use.names=TRUE, idcol="Number of genes in Set", fill=TRUE)
write.csv(complete.sim.results,file=paste0(N,"_replicates_all_GSEA_ties_results_simulation_",Sys.Date(),".csv"), row.names=FALSE)
write.csv(complete.synth.sim.results,file=paste0(N,"_replicates_synthetic_gene_Set_GSEA_ties_results_",Sys.Date(),".csv"), row.names=FALSE)
complete.perc.sig <- data.table::rbindlist(all.perc.sig, use.names=TRUE, idcol="Number of genes in Set", fill=TRUE)
write.csv(complete.perc.sig,file=paste0(N,"_replicates_percent_significant_synthetic_gene_set_enrichments_",Sys.Date(),".csv"), row.names=FALSE)
## Generate tile plots
# prepare collapsed results
og.complete.perc.sig$N_genes <- as.numeric(og.complete.perc.sig$`Number of genes in Set`)
og.complete.perc.sig$values.to.vary <- as.numeric(og.complete.perc.sig$values.to.vary)
complete.perc.sig$N_genes <- as.numeric(complete.perc.sig$`Number of genes in Set`)
complete.perc.sig$values.to.vary <- as.numeric(complete.perc.sig$values.to.vary)
# load theme for plot
ng.theme <- theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
panel.border = element_rect(fill=NA), panel.background = element_blank(),
axis.line = element_line(colour = "black"), axis.text.x = element_text(colour = "black"),
axis.text.y = element_text(colour = "black"), axis.ticks.x = element_line(colour="black"),
axis.ticks.y = element_line(colour="black"), legend.title = element_blank(),
axis.title.y = element_text(size=8, colour="black"))
# produce tile plots
cutoff <- c(0.05, 0.25)
for (i in cutoff) {
temp.perc.col <- ifelse(i == 0.05, "perc.moreSig", "perc.sig")
temp.cols <- c("N_genes", "values.to.vary", temp.perc.col)
# original order
temp.og <- og.complete.perc.sig[,..temp.cols]
colnames(temp.og)[3] <- "perc"
q.tile.plot <- ggplot(temp.og, aes(x = as.factor(N_genes), y = values.to.vary, fill = perc)) +
geom_tile(height=0.25) + ng.theme + theme_light() + scale_fill_gradient2() +
scale_x_discrete(breaks = gene.set.sizes) +
labs(x="Number of Genes in Set", y = "Value Added", fill = paste0("% q < ", i))
ggsave(q.tile.plot, file = paste0("GSEA_ties_sim_p0.05_q",i,"_N",N,"_originalOrder_",Sys.Date(),".pdf"))
# shuffled ties
temp.df <- complete.perc.sig[,..temp.cols]
colnames(temp.df)[3] <- "perc"
q.tile.plot <- ggplot(temp.df, aes(x = as.factor(N_genes), y = values.to.vary, fill = perc)) +
geom_tile(height=0.25) + ng.theme + theme_light() + scale_fill_gradient2() +
scale_x_discrete(breaks = gene.set.sizes) +
labs(x="Number of Genes in Set", y = "Value Added", fill = paste0("% q < ", i))
ggsave(q.tile.plot, file = paste0("GSEA_ties_sim_p0.05_q",i,"_N",N,"_",Sys.Date(),".pdf"))
}
NES.tile.plot <- ggplot(og.complete.perc.sig, aes(x = as.factor(N_genes), y = values.to.vary, fill=avg.NES)) +
geom_tile(height=0.25) + ng.theme + theme_light() + scale_fill_gradient2() +
scale_x_discrete(breaks = gene.set.sizes) +
labs(x="Number of Genes in Set", y = "Value Added", fill = "Mean NES")
ggsave(NES.tile.plot, file = paste0("GSEA_ties_sim_NES_N",N,"_originalOrder_",Sys.Date(),".pdf"))
NES.tile.plot <- ggplot(complete.perc.sig, aes(x = as.factor(N_genes), y = values.to.vary, fill=avg.NES)) +
geom_tile(height=0.25) + ng.theme + theme_light() + scale_fill_gradient2() +
scale_x_discrete(breaks = gene.set.sizes) +
labs(x="Number of Genes in Set", y = "Value Added", fill = "Mean NES")
ggsave(NES.tile.plot, file = paste0("GSEA_ties_sim_NES_N",N,"_",Sys.Date(),".pdf"))