/
3_Enrichment_HomVsHet.Rmd
499 lines (424 loc) · 13.7 KB
/
3_Enrichment_HomVsHet.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
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
---
title: "Enrichment Analysis: Homozygous Vs Heterozygous Mutants"
author: "Steve Pederson"
date: "`r format(Sys.time(), '%d %B, %Y')`"
output: workflowr::wflow_html
editor_options:
chunk_output_type: console
---
```{r setup, include=FALSE}
knitr::opts_chunk$set(
autodep = TRUE,
echo = TRUE,
warning = FALSE,
message = FALSE,
fig.align = "center"
)
```
# Setup
```{r loadPackages}
library(tidyverse)
library(magrittr)
library(edgeR)
library(scales)
library(pander)
library(msigdbr)
library(AnnotationDbi)
library(RColorBrewer)
library(ngsReports)
library(fgsea)
library(metap)
```
```{r setOpts}
theme_set(theme_bw())
panderOptions("table.split.table", Inf)
panderOptions("table.style", "rmarkdown")
panderOptions("big.mark", ",")
```
```{r samplesAndLabels}
samples <- here::here("data/samples.csv") %>%
read_csv() %>%
distinct(sampleName, .keep_all = TRUE) %>%
dplyr::select(sample = sampleName, sampleID, genotype) %>%
mutate(
genotype = factor(genotype, levels = c("WT", "Het", "Hom")),
mutant = genotype %in% c("Het", "Hom"),
homozygous = genotype == "Hom"
)
genoCols <- samples$genotype %>%
levels() %>%
length() %>%
brewer.pal("Set1") %>%
setNames(levels(samples$genotype))
```
```{r loadFits}
dgeList <- here::here("data/dgeList.rds") %>% read_rds()
cpmPostNorm <- here::here("data/cpmPostNorm.rds") %>% read_rds()
entrezGenes <- dgeList$genes %>%
dplyr::filter(!is.na(entrezid)) %>%
unnest(entrezid) %>%
dplyr::rename(entrez_gene = entrezid)
```
```{r topTables}
deTable <- here::here("output", "psen2HomVsHet.csv") %>%
read_csv() %>%
mutate(
entrezid = dgeList$genes$entrezid[gene_id]
)
```
```{r formatP}
formatP <- function(p, m = 0.0001){
out <- rep("", length(p))
out[p < m] <- sprintf("%.2e", p[p<m])
out[p >= m] <- sprintf("%.4f", p[p>=m])
out
}
```
# Introduction
Enrichment analysis for this dataset present some challenges.
Despite normalisation to account for gene length and GC bias, some appeared to still be present in the final results.
In addition, the confounding of incomplete rRNA removal with genotype may lead to other distortions in both DE genes and ranking statistics.
As the list of DE genes for this comparison was small ($n_{\text{DE}} = `r nrow(dplyr::filter(deTable, DE))`$), enrichment testing was only performed using ranked-list approaches.
Testing for enrichment *with ranked lists* will be performed using:
1. `fry` as this can take into account inter-gene correlations. Values supplied will be fitted values for each gene/sample as these have been corrected for GC and length biases.
2. `camera`, which also accommodates inter-gene correlations. Values supplied will be fitted values for each gene/sample as these have been corrected for GC and length biases.
3. `fgsea` which is an R implementation of GSEA. This approach simply takes a ranked list and doesn't directly account for correlations. However, the ranked list will be derived from analysis using CQN-normalisation so will be robust to these technical artefacts.
In the case of `camera`, inter-gene correlations will be calculated for each gene-set prior to analysis to ensure more conservative p-values are obtained.
# Databases used for testing
Data was sourced using the `msigdbr` package.
The initial database used for testing was the Hallmark Gene Sets, with mappings from gene-set to EntrezGene IDs performed by the package authors.
## Hallmark Gene Sets
```{r hm}
hm <- msigdbr("Danio rerio", category = "H") %>%
left_join(entrezGenes) %>%
dplyr::filter(!is.na(gene_id)) %>%
distinct(gs_name, gene_id, .keep_all = TRUE)
hmByGene <- hm %>%
split(f = .$gene_id) %>%
lapply(extract2, "gs_name")
hmByID <- hm %>%
split(f = .$gs_name) %>%
lapply(extract2, "gene_id")
```
Mappings are required from gene to pathway, and Ensembl identifiers were used to map from gene to pathway, based on the mappings in the previously used annotations (Ensembl Release 98).
A total of `r comma(length(hmByGene))` Ensembl IDs were mapped to pathways from the hallmark gene sets.
## KEGG Gene Sets
```{r kg}
kg <- msigdbr("Danio rerio", category = "C2", subcategory = "CP:KEGG") %>%
left_join(entrezGenes) %>%
dplyr::filter(!is.na(gene_id)) %>%
distinct(gs_name, gene_id, .keep_all = TRUE)
kgByGene <- kg %>%
split(f = .$gene_id) %>%
lapply(extract2, "gs_name")
kgByID <- kg %>%
split(f = .$gs_name) %>%
lapply(extract2, "gene_id")
```
The same mapping process was applied to KEGG gene sets.
A total of `r comma(length(kgByGene))` Ensembl IDs were mapped to pathways from the KEGG gene sets.
## Gene Ontology Gene Sets
```{r goSummaries}
goSummaries <- url("https://uofabioinformaticshub.github.io/summaries2GO/data/goSummaries.RDS") %>%
readRDS() %>%
mutate(
Term = Term(id),
gs_name = Term %>% str_to_upper() %>% str_replace_all("[ -]", "_"),
gs_name = paste0("GO_", gs_name)
)
minPath <- 3
```
```{r go}
go <- msigdbr("Danio rerio", category = "C5") %>%
left_join(entrezGenes) %>%
dplyr::filter(!is.na(gene_id)) %>%
left_join(goSummaries) %>%
dplyr::filter(shortest_path >= minPath) %>%
distinct(gs_name, gene_id, .keep_all = TRUE)
goByGene <- go %>%
split(f = .$gene_id) %>%
lapply(extract2, "gs_name")
goByID <- go %>%
split(f = .$gs_name) %>%
lapply(extract2, "gene_id")
```
For analysis of gene-sets from the GO database, gene-sets were restricted to those with `r minPath` or more steps back to the ontology root terms.
A total of `r comma(length(goByGene))` Ensembl IDs were mapped to pathways from restricted database of `r comma(length(goByID))` GO gene sets.
```{r gsSizes}
gsSizes <- bind_rows(hm, kg, go) %>%
dplyr::select(gs_name, gene_symbol, gene_id) %>%
chop(c(gene_symbol, gene_id)) %>%
mutate(gs_size = vapply(gene_symbol, length, integer(1)))
```
# Enrichment Testing on Ranked Lists
```{r rnk}
rnk <- structure(
-sign(deTable$logFC)*log10(deTable$PValue),
names = deTable$gene_id
) %>% sort()
np <- 1e5
```
Genes were ranked by -sign(logFC)*log~10~(p) for approaches which required a ranked list.
Multiple approaches were first calculated individually, before being combined for the final integrated set of results.
## Hallmark Gene Sets
```{r hmFry}
hmFry <- cpmPostNorm %>%
fry(
index = hmByID,
design = dgeList$design,
contrast = "homozygous",
sort = "directional"
) %>%
rownames_to_column("gs_name") %>%
as_tibble()
```
For analysis under `camera` when inter-gene correlations were calculated for a more conservative result.
```{r hmCamera}
hmCamera <- cpmPostNorm %>%
camera(
index = hmByID,
design = dgeList$design,
contrast = "homozygous",
inter.gene.cor = NULL
) %>%
rownames_to_column("gs_name") %>%
as_tibble()
```
For generation of the GSEA ranked list, `r comma(np)` permutations were conducted.
```{r hmGsea}
hmGsea <- fgsea(
pathways = hmByID,
stats = rnk,
nperm = np
) %>%
as_tibble() %>%
dplyr::rename(gs_name = pathway, PValue = pval) %>%
arrange(PValue)
```
Results for all analyses were then combined using Wilkinson's method to combine p-values.
For a conservative approach, under $m$ tests, the $m - 1^{\text{th}}$ smallest p-value was chosen.
```{r hmMeta}
hmMeta <- hmFry %>%
dplyr::select(gs_name, fry = PValue) %>%
left_join(
dplyr::select(hmCamera, gs_name, camera = PValue)
) %>%
left_join(
dplyr::select(hmGsea, gs_name, gsea = PValue)
) %>%
nest(p = one_of(c("fry", "camera", "gsea"))) %>%
mutate(
n_p = vapply(p, function(x){sum(!is.na(unlist(x)))}, integer(1)),
wilkinson_p = vapply(p, function(x){
x <- unlist(x)
x <- x[!is.na(x)]
wilkinsonp(x, length(x) - 1)$p
}, numeric(1)),
FDR = p.adjust(wilkinson_p, "fdr"),
adjP = p.adjust(wilkinson_p, "bonferroni")
) %>%
arrange(wilkinson_p) %>%
unnest(p) %>%
left_join(gsSizes) %>%
mutate(
DE = lapply(gene_id, intersect, dplyr::filter(deTable, DE)$gene_id),
DE = lapply(DE, unique),
nDE = vapply(DE, length, integer(1))
)
```
```{r}
hmMeta %>%
dplyr::filter(FDR < 0.1) %>%
mutate_at(vars(one_of(c("wilkinson_p", "FDR", "adjP"))), formatP) %>%
dplyr::select(`Gene Set` = gs_name, `Number DE` = nDE, `Set Size` = gs_size, `Wilkinson~p~` = wilkinson_p, `p~FDR~` = FDR, `p~bonf~` = adjP) %>%
pander(
caption = "Results from combining all above approaches for the Hallmark Gene Sets. All terms are significant to an FDR of 0.1, with none passing the initial filter of FDR < 0.05",
justify = "lrrrrr"
)
```
## KEGG Gene Sets
```{r kgFry}
kgFry <-cpmPostNorm%>%
fry(
index = kgByID,
design = dgeList$design,
contrast = "homozygous",
sort = "directional"
) %>%
rownames_to_column("gs_name") %>%
as_tibble()
```
For analysis under `camera` when inter-gene correlations were calculated for a more conservative result.
```{r kgCamera}
kgCamera <- cpmPostNorm %>%
camera(
index = kgByID,
design = dgeList$design,
contrast = "homozygous",
inter.gene.cor = NULL
) %>%
rownames_to_column("gs_name") %>%
as_tibble()
```
For generation of the GSEA ranked list, `r comma(np)` permutations were conducted.
```{r kgGsea}
kgGsea <- fgsea(
pathways = kgByID,
stats = rnk,
nperm = np
) %>%
as_tibble() %>%
dplyr::rename(gs_name = pathway, PValue = pval) %>%
arrange(PValue)
```
Results for all analyses were then combined using Wilkinson's method to combine p-values.
For a conservative approach, under $m$ tests, the $m - 1^{\text{th}}$ smallest p-value was chosen.
```{r kgMeta}
kgMeta <- kgFry %>%
dplyr::select(gs_name, fry = PValue) %>%
left_join(
dplyr::select(kgCamera, gs_name, camera = PValue)
) %>%
left_join(
dplyr::select(kgGsea, gs_name, gsea = PValue)
) %>%
nest(p = one_of(c("fry", "camera", "gsea"))) %>%
mutate(
n_p = vapply(p, function(x){sum(!is.na(unlist(x)))}, integer(1)),
wilkinson_p = vapply(p, function(x){
x <- unlist(x)
x <- x[!is.na(x)]
wilkinsonp(x, length(x) - 1)$p
}, numeric(1)),
FDR = p.adjust(wilkinson_p, "fdr"),
adjP = p.adjust(wilkinson_p, "bonferroni")
) %>%
arrange(wilkinson_p) %>%
unnest(p) %>%
left_join(gsSizes) %>%
mutate(
DE = lapply(gene_id, intersect, dplyr::filter(deTable, DE)$gene_id),
DE = lapply(DE, unique),
nDE = vapply(DE, length, integer(1))
)
```
```{r}
kgMeta %>%
dplyr::filter(FDR < 0.01) %>%
mutate_at(vars(one_of(c("wilkinson_p", "FDR", "adjP"))), formatP) %>%
dplyr::select(`Gene Set` = gs_name, `Number DE` = nDE, `Set Size` = gs_size, `Wilkinson~p~` = wilkinson_p, `p~FDR~` = FDR, `p~bonf~` = adjP) %>%
pander(
caption = "Results from combining all above approaches for the KEGG Gene Sets. All terms are significant to an FDR of 0.05.",
justify = "lrrrrr"
)
```
## GO Gene Sets
```{r goFry}
goFry <- cpmPostNorm %>%
fry(
index = goByID,
design = dgeList$design,
contrast = "homozygous",
sort = "directional"
) %>%
rownames_to_column("gs_name") %>%
as_tibble()
```
For analysis under `camera` when inter-gene correlations were calculated for a more conservative result.
```{r goCamera}
goCamera <- cpmPostNorm %>%
camera(
index = goByID,
design = dgeList$design,
contrast = "homozygous",
inter.gene.cor = NULL
) %>%
rownames_to_column("gs_name") %>%
as_tibble()
```
For generation of the GSEA ranked list, `r comma(np)` permutations were conducted.
```{r goGsea}
goGsea <- fgsea(
pathways = goByID,
stats = rnk,
nperm = np
) %>%
as_tibble() %>%
dplyr::rename(gs_name = pathway, PValue = pval) %>%
arrange(PValue)
```
Results for all analyses were then combined using Wilkinson's method to combine p-values.
For a conservative approach, under $m$ tests, the $m - 1^{\text{th}}$ smallest p-value was chosen.
```{r goMeta}
goMeta <- goFry %>%
dplyr::select(gs_name, fry = PValue) %>%
left_join(
dplyr::select(goCamera, gs_name, camera = PValue)
) %>%
left_join(
dplyr::select(goGsea, gs_name, gsea = PValue)
) %>%
nest(p = one_of(c("fry", "camera", "gsea"))) %>%
mutate(
n_p = vapply(p, function(x){sum(!is.na(unlist(x)))}, integer(1)),
wilkinson_p = vapply(p, function(x){
x <- unlist(x)
x <- x[!is.na(x)]
wilkinsonp(x, length(x) - 1)$p
}, numeric(1)),
FDR = p.adjust(wilkinson_p, "fdr"),
adjP = p.adjust(wilkinson_p, "bonferroni")
) %>%
arrange(wilkinson_p) %>%
unnest(p) %>%
left_join(gsSizes) %>%
mutate(
DE = lapply(gene_id, intersect, dplyr::filter(deTable, DE)$gene_id),
DE = lapply(DE, unique),
nDE = vapply(DE, length, integer(1))
)
```
```{r}
goMeta %>%
dplyr::filter(adjP < 0.05) %>%
mutate_at(vars(one_of(c("wilkinson_p", "FDR", "adjP"))), formatP) %>%
dplyr::select(`Gene Set` = gs_name, `Number DE` = nDE, `Set Size` = gs_size, `Wilkinson~p~` = wilkinson_p, `p~FDR~` = FDR, `p~bonf~` = adjP) %>%
pander(
caption = "Results from combining all above approaches for the GO Gene Sets. All terms are significant to a bonferroni-adjusted p-value < 0.05.",
justify = "lrrrrr"
)
```
# Data Export
All enriched gene sets terms with an FDR adjusted p-value < 0.05 were exported as a single csv file.
```{r}
add_prefix <- function(x, pre = "p_"){
paste0(pre, x)
}
bind_rows(
hmMeta,
kgMeta,
goMeta
) %>%
dplyr::filter(FDR < 0.05) %>%
mutate(
DE = lapply(DE, function(x){dplyr::filter(deTable, gene_id %in% x)$gene_name}),
DE = lapply(DE, unique),
DE = vapply(DE, paste, character(1), collapse = ";")
) %>%
arrange(wilkinson_p) %>%
dplyr::select(
`Gene Set` = gs_name,
`Nbr Detected Genes` = gs_size,
`Nbr DE Genes` = nDE,
combined = wilkinson_p, FDR,
fry, camera, gsea,
`DE Genes` = DE
) %>%
rename_at(
vars(combined, fry, camera, gsea),
add_prefix
) %>%
write_csv(
here::here("output", "Enrichment_Hom_V_Het.csv")
)
```