-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathgsea_report_template.Rmd
275 lines (212 loc) · 8.46 KB
/
gsea_report_template.Rmd
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
260
261
262
263
264
265
266
267
268
269
270
271
272
273
---
title: "GSEA"
author: "Kimon"
date: "10/14/2022"
output:
html_document:
code_folding: hide
toc: true
toc_float:
collapsed: false
params:
selfname: "gsea.html"
resultsDir: "./results/GSEA"
gsdb: "/Volumes/groups/busslinger/Bioinf/src/cbe/broad/msigdb_v7.5.1_files_to_download_locally/msigdb_v7.5.1_GMTs/c5.go.mf.v7.5.1.symbols.gmt"
xorg: "/Volumes/groups/busslinger/Bioinf/src/cbe/broad/Mouse_Gene_Symbol_Remapping_Human_Orthologs_MSigDB.v7.5.1.chip"
de: '/Volumes/groups/busslinger/Kimon/tanja/R13870_RNAseq_timecourse/results/DE/exon_genecounts.CellType_actB.Condition_Ikaros_00_5phIAA_vs_Aiolos_00_5phIAA.deseq2.nolab.spotfire.txt'
alpha: 0.05
fx: "log2FoldChange"
p: NULL
pcut: 0.05
---
# Setup
```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
library(fgsea)
library(grid)
library(ggplot2)
library(data.table)
message(paste("DE file:", params$de))
message(paste("Gene-set file:", params$gsdb))
message(paste("Alpha:", params$alpha))
message(paste("Ranking values:", params$fx))
```
```{r prepare}
# Import gene sets
GSDB <- fgsea::gmtPathways(params$gsdb)
# DESeq2 data (with shrunken LFCs)
DE <- fread(params$de)
DE <- DE[!is.na(DE[[params$fx]]), ]
if (!is.null(params$p))
DE <- DE[!is.na(DE[[params$p]]) & DE[[params$p]] < params$pcut, ]
setorderv(DE, cols=params$fx, order=-1)
genes.rnk <- DE[[params$fx]]
message(paste("Number of genes:", length(genes.rnk)))
if (!is.null(params$xorg)) {
# Convert gene names to species orthologs
XREF <- fread(params$xorg)
setnames(XREF, old=names(XREF)[1:2], new=c('from', 'to'))
setkey(XREF, from)
names(genes.rnk) <- XREF[DE$name, to]
names(genes.rnk)[is.na(names(genes.rnk))] <- DE$name[is.na(names(genes.rnk))]
} else {
names(genes.rnk) <- DE$name
}
```
# GSEA
Genes are ordered and ranked by shrunken log-fold-change. Genes that are at a tie (same shrunken LFC) will be ranked in arbitrary order, which may affect the order of some gene-sets relative to one another.
```{r gsea, warning=FALSE}
# Wrapper for {fgsea}, https://bioinformaticsbreakdown.com/how-to-gsea/
GSEA = function(gene_list, myGO, pval=params$alpha) {
set.seed(54321)
library(dplyr)
library(fgsea)
if ( any( duplicated(names(gene_list)) ) ) {
warning("Duplicates in gene names")
gene_list = gene_list[!duplicated(names(gene_list))]
}
if ( !all( order(gene_list, decreasing = TRUE) == 1:length(gene_list)) ){
warning("Gene list not sorted")
gene_list = sort(gene_list, decreasing = TRUE)
}
fgRes <- fgsea::fgsea(pathways = myGO,
stats = gene_list,
minSize=15, ## minimum gene set size
maxSize=400 ) %>% # , ## maximum gene set size
# nperm=10000) %>%
as.data.frame() %>%
dplyr::filter(padj < !!pval) %>%
arrange(desc(NES))
message(paste("Number of signficant gene sets =", nrow(fgRes)))
message("Collapsing Pathways -----")
concise_pathways = collapsePathways(data.table::as.data.table(fgRes),
pathways = myGO,
stats = gene_list)
fgRes = fgRes[fgRes$pathway %in% concise_pathways$mainPathways, ]
message(paste("Number of gene sets after collapsing =", nrow(fgRes)))
fgRes$Enrichment = ifelse(fgRes$NES > 0, "Up-regulated", "Down-regulated")
filtRes = rbind(head(fgRes, n = 10),
tail(fgRes, n = 10 ))
total_up = sum(fgRes$Enrichment == "Up-regulated")
total_down = sum(fgRes$Enrichment == "Down-regulated")
header = paste0("Top 10 (Total pathways: Up=", total_up,", Down=", total_down, ")")
colos = setNames(c("firebrick2", "dodgerblue2"),
c("Up-regulated", "Down-regulated"))
g1= ggplot(filtRes, aes(reorder(pathway, NES), NES)) +
geom_point( aes(fill = Enrichment, size = size), shape=21) +
scale_fill_manual(values = colos ) +
scale_size_continuous(range = c(2,10)) +
geom_hline(yintercept = 0) +
coord_flip() +
labs(x="Pathway", y="Normalized Enrichment Score",
title=header)
output = list("Results" = fgRes, "Plot" = g1)
return(output)
}
# GSEA
res <- GSEA(genes.rnk, GSDB)
res$Results <- as.data.table(res$Results)
```
## Top gene sets
Top 10 each enriched and depleted gene-sets.
```{r sets, fig.height=10, fig.width=10}
print(res$Plot)
```
## Clustering of overlapping gene sets
Cluster gene-sets based on shared genes.
```{r clusters}
# Cluster gene sets, https://bioinformaticsbreakdown.com/how-to-gsea/
plot_geneset_clusters = function(gs_results, myGO, min.sz = 4, main="GSEA clusters"){
library(ggplot2)
library(ggrepel)
library(stringr)
df = matrix(nrow=nrow(gs_results), ncol = nrow(gs_results), data = 0)
rownames(df) = colnames(df) = gs_results$pathway
for ( i in 1:nrow(gs_results)) {
genesI = unlist(myGO[names(myGO) == gs_results$pathway[i] ])
for (j in 1:nrow(gs_results)) {
genesJ = unlist(myGO[names(myGO) == gs_results$pathway[j] ])
## Jaccards distance 1 - (intersection / union )
overlap = sum(!is.na(match(genesI, genesJ )))
jaccards = overlap / length(unique(c(genesI, genesJ) ))
df[i,j] = 1-jaccards
}
}
## Cluster nodes using dynamic tree cut, for colors
distMat = as.dist(df)
dendro = hclust(distMat, method = "average" )
clust = dynamicTreeCut::cutreeDynamicTree( dendro, minModuleSize = min.sz )
## Note: in dynamicTreeCut, cluster 0, is a garbage cluster for things that dont cluster, so we remove it
gs_results$Cluster = clust
gs_results = gs_results[gs_results$Cluster != 0, ]
## select gene sets to label for each clusters
bests = gs_results %>%
group_by( Cluster ) %>%
top_n(wt = abs(size), n = 1) %>%
.$pathway
## determine cluster order for plotting
clust_ords = gs_results %>%
group_by( Cluster ) %>%
summarise("Average" = NES ) %>%
arrange(desc(Average)) %>%
.$Cluster %>%
unique
gs_results$Cluster = factor(gs_results$Cluster, levels = clust_ords)
gs_results$Label = ""
gs_results$Label[gs_results$pathway %in% bests ] = gs_results$pathway[gs_results$pathway %in% bests ]
gs_results$Label = str_remove(gs_results$Label, "GO_")
gs_results$Label = tolower(gs_results$Label)
g1 = ggplot(gs_results, aes(x = Cluster, y = NES, label = Label )) +
geom_jitter( aes(color = Cluster, size = size), alpha = 0.8, height = 0, width = 0.2 ) +
scale_size_continuous(range = c(0.5,5)) +
geom_text_repel( force = 2, max.overlaps = Inf) +
ggtitle(main)
return(g1)
}
print( plot_geneset_clusters( gs_results = res$Results[res$Results$NES > 0, ], GSDB,
main = "Up-regulated GSEA clusters (min 4)",
min.sz = min(4, nrow(res$Results[res$Results$NES > 0, ]) - 1) ) )
print( plot_geneset_clusters( gs_results = res$Results[res$Results$NES > 0, ], GSDB,
main = "Up-regulated GSEA clusters (min 2)",
min.sz = min(2, nrow(res$Results[res$Results$NES > 0, ]) - 1) ) )
print( plot_geneset_clusters( gs_results = res$Results[res$Results$NES < 0, ], GSDB,
main = "Down-regulated GSEA clusters (min 4)",
min.sz = min(4, nrow(res$Results[res$Results$NES < 0, ]) - 1) ) )
print( plot_geneset_clusters( gs_results = res$Results[res$Results$NES < 0, ], GSDB,
main = "Down-regulated GSEA clusters (min 2)",
min.sz = min(2, nrow(res$Results[res$Results$NES < 0, ]) - 1) ) )
```
## Gene ranks in the top gene-sets
### All enriched sets
```{r ranks1, fig.height=15, fig.width=10}
# Table of enrichment plots
topPathwaysUp <- res$Results[ES > 0, ][head(order(pval), n=25), pathway]
topPathwaysDown <- res$Results[ES < 0, ][head(order(pval), n=25), pathway]
topPathways <- c(topPathwaysUp, rev(topPathwaysDown))
grid.newpage(); par(mfg=c(1,1)) # Prevent overlay with whatever previous plot
print( plotGseaTable(GSDB[topPathways], genes.rnk, res$Results, gseaParam=0.5) )
```
<!-- ### Collapsed overlaps -->
<!-- ```{r ranks2, fig.height=10} -->
<!-- # Select only independent pathways -->
<!-- collapsedPathways <- collapsePathways(res$Results[order(pval)][padj < 0.01], GSDB, genes.rnk) -->
<!-- mainPathways <- res$Results[pathway %in% collapsedPathways$mainPathways][order(-NES), pathway] -->
<!-- grid.newpage(); par(mfg=c(1,1)) # Prevent overlay with whatever previous plot -->
<!-- print( plotGseaTable(GSDB[mainPathways], genes.rnk, res$Results, gseaParam = 0.3) ) -->
<!-- ``` -->
### Enrichment per set
```{r ranks3, fig.height=2.5, fig.width=8}
for(gs in res$Results$pathway) {
# Enrichment plot of one cluster
print( plotEnrichment(GSDB[[gs]], genes.rnk) +
labs(title=gs) )
}
```
# File out
```{r out}
fwrite(res$Results, file=file.path(params$resultsDir, sub('html$', 'txt', params$selfname)))
```
# Session info
```{r}
sessionInfo()
```