-
Notifications
You must be signed in to change notification settings - Fork 0
/
helD.Rmd
247 lines (183 loc) · 8.47 KB
/
helD.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
---
title: "helD alternative analysis"
output: html_notebook
---
```{r setup}
library(DESeq2)
library(here)
library(tidyverse)
includeExp1 = TRUE
```
# Load data
Based largerly on the original code by Martin Převorovský (Thanks!)
## Read first experiment
The first experiment has extra deletion AND a lots of rRNA reads :-/
It also compares SPOrulation to EXPonential and uses a different medium than the other two.
```{r}
# prepare summarized experiment metadata
samples1 <- c('helD_EXP_3', 'WT_EXP_1', 'WT_SPO_3', 'helD_SPO_3', 'helD_EXP_1', 'WT_SPO_1', 'helD_SPO_1', 'WT_EXP_2', 'helD_EXP_2', 'WT_SPO_2', 'helD_SPO_2', 'WT_EXP_3') # NOTE the seemingly irregular order of samples is due alphabetical sorting of file names containing number (810, 81, 811, 822, 82...)
sampleInfo1 <- data.frame(t(read.csv(here('private_data','sampleInfo'), sep='\t', head=F)))
colnames(sampleInfo1) <- c('run', 'genotype', 'file', 'phase')
sampleInfo1 <- sampleInfo1 %>% mutate(sample = samples1, experiment = "1",
extra_deletion = (genotype == "helD_KO"),
growth_medium = "A")
counts1 <- read.csv(here('private_data','counts.txt'))
colnames(counts1)[1] <- "gene"
# remove contaminated samples (only ~5% reads mapped)
sampleInfo1 <- sampleInfo1 %>% filter(!(sample %in% c("helD_EXP_3","helD_SPO_3")))
counts1 <- counts1 %>% select(-helD_EXP_3, -helD_SPO_3)
if(!all(colnames(counts1)[2:length(counts1)] == sampleInfo1$sample)) {
stop("Mismatched count columns and sample info")
}
colnames(counts1)[2:length(counts1)] <- paste0("Exp1_",colnames(counts1)[2:length(counts1)])
```
## Read second experiment
This experiment still had extra deletion with helD
```{r}
# prepare summarized experiment metadata
samples2 <- c('helD_1', 'helD_2', 'helD_3', 'WT_1', 'WT_2', 'WT_3')
sampleInfo2 <- data.frame(t(read.csv(here('private_data','sampleInfo_r2'), sep = '\t', head = FALSE)))
colnames(sampleInfo2) <- c('run', 'genotype', 'file')
sampleInfo2 <- sampleInfo2 %>% mutate(sample = samples2, phase = 'EXP', experiment = "2",
extra_deletion = (genotype == "helD_KO"),
growth_medium = "B")
counts2 <- read.csv(here('private_data','counts_r2.txt'))
colnames(counts2)[1] <- "gene"
if(!all(colnames(counts2)[2:length(counts2)] == samples2)) {
stop("Mismatched count columns and sample info")
}
colnames(counts2)[2:length(counts2)] <- paste0("Exp2_",colnames(counts2)[2:length(counts2)])
```
## Read third experiment
```{r}
samples3 <- c('helD_EXP_A', 'helD_STAT_A', 'helD_EXP_B', 'helD_STAT_B', 'helD_EXP_D', 'helD_STAT_D',
'WT_EXP_A', 'WT_STAT_A', 'WT_EXP_B', 'WT_STAT_B', 'WT_EXP_D', 'WT_STAT_D')
sampleInfo3 <- data.frame(t(read.csv(here('private_data','sampleInfo_r3'), sep = '\t', head = FALSE)))
colnames(sampleInfo3) <- c('genotype', 'phase', 'run', 'file')
sampleInfo3 <- sampleInfo3 %>% mutate(sample = samples3, experiment = "3",
extra_deletion = FALSE, growth_medium = "B",
genotype = fct_recode(genotype, "helD_KO" = "helD KO"))
counts3 <- read.csv(here('private_data','counts_r3.txt'))
colnames(counts3)[1] <- "gene"
# remove sample that failed QC (mixed EXP and STAT?)
sampleInfo3 <- sampleInfo3 %>% filter(sample != "WT_EXP_A")
counts3 <- counts3 %>% select(-WT_EXP_A)
if(!all(colnames(counts3)[2:length(counts3)] == sampleInfo3$sample)) {
stop("Mismatched count columns and sample info")
}
colnames(counts3)[2:length(counts3)] <- paste0("Exp3_",colnames(counts3)[2:length(counts3)])
```
# Check gene overlap in the data
```{r}
num_mismatch = 0
num_mismatch <- num_mismatch + length(setdiff(counts1$gene, counts2$gene))
num_mismatch <- num_mismatch + length(setdiff(counts1$gene, counts3$gene))
num_mismatch <- num_mismatch + length(setdiff(counts2$gene, counts1$gene))
num_mismatch <- num_mismatch + length(setdiff(counts2$gene, counts3$gene))
num_mismatch <- num_mismatch + length(setdiff(counts3$gene, counts1$gene))
num_mismatch <- num_mismatch + length(setdiff(counts3$gene, counts2$gene))
if(num_mismatch > 0) {
stop("Not all genes are in all count files")
}
```
# Create joint dataset
I am filtering out rRNA counts (expecially because of the rRNA problem in Exp1)
```{r}
counts_all_matrix <-
{
if(includeExp1) { counts1 }
else { counts1 %>% head(0) }
} %>%
inner_join(counts2, c("gene" = "gene")) %>%
inner_join(counts3, c("gene" = "gene")) %>%
filter(!grepl("^BSU_rRNA_", gene)) %>%
column_to_rownames("gene") %>% as.matrix()
sampleInfo_all <- rbind(
if(includeExp1) { sampleInfo1 }
else { sampleInfo1 %>% head(0) },
sampleInfo2, sampleInfo3
) %>%
mutate(run = factor(paste0("Exp",experiment,"_",run)),
sample = factor(paste0("Exp",experiment,"_",sample)),
genotype = factor(genotype, levels = c("WT","helD_KO")), #relevel(genotype, ref = "WT"),
phase = factor(phase, levels = c("STAT","EXP","SPO")), # relevel(phase, ref = "STAT"),
extra_deletion = factor(extra_deletion, levels = c(FALSE, TRUE), labels = c("no", "yes")),
genotype_phase = interaction(genotype, phase)
) %>%
droplevels()
if(!all(sampleInfo_all$sample == colnames(counts_all_matrix))) {
stop("Mismatched count columns and sample info")
}
```
# The actual DESeq
Using shrinkage - the `betaPrior` argument (which is not default), as without shrinkage there are convergence issues.
```{r}
# Currently We ignore growth medium as this is covered by the run variable
dds <- DESeqDataSetFromMatrix(countData = counts_all_matrix,
colData = sampleInfo_all,
design = ~ genotype + extra_deletion + run + phase)
#design = ~ genotype_phase + extra_deletion + run)
#dds <- DESeq(dds, betaPrior = TRUE)
dds <- estimateSizeFactors(dds)
dds <- estimateDispersions(dds)
dds <- nbinomWaldTest(dds, betaPrior = TRUE)
```
```{r}
res_helD <- results(dds, name = "helD_KO", contrast = c("genotype","helD_KO","WT"))
#res_helD_lfc <- lfcShrink(dds)
res_helD %>% as.data.frame() %>% rownames_to_column("gene") %>% filter(gene == "BSU33450")#filter(padj <= 0.05)
plotMA(res_helD)
plotCounts(dds, gene = "BSU33450", intgroup = "genotype", normalized = TRUE )
plotCounts(dds, gene = "BSU33450", intgroup = "phase", normalized = TRUE )
```
```{r}
res_extra_del <- results(dds, name = "extra_deletion", contrast = c("extra_deletion","yes","no"))
#res_helD_lfc <- lfcShrink(dds)
res_extra_del %>% as.data.frame() %>% rownames_to_column("gene") %>% filter(padj <= 0.05)
plotCounts(dds, gene = "BSU04250", intgroup = "extra_deletion", normalized = FALSE )
```
```{r}
res_exp_phase <- results(dds, name = "exp_vs_stat", contrast = c("phase","EXP","STAT"))
#res_helD_lfc <- lfcShrink(dds)
res_exp_phase %>% as.data.frame() %>% rownames_to_column("gene") %>% filter(padj <= 0.05)
res_exp_phase %>% as.data.frame() %>% rownames_to_column("gene") %>%filter(gene == "BSU33450")
```
```{r}
run_diffs_list <- list()
for(run_index_1 in 1:(length(levels(sampleInfo_all$run)) - 1)) {
for(run_index_2 in (run_index_1 + 1):length(levels(sampleInfo_all$run))) {
run1 = levels(sampleInfo_all$run)[run_index_1]
run2 = levels(sampleInfo_all$run)[run_index_2]
res_run <- results(dds, name = "run_test", contrast = c("run",run1,run2))
#res_helD_lfc <- lfcShrink(dds)
n_diff <- res_run %>% as.data.frame() %>% rownames_to_column("gene") %>% filter(padj <= 0.05) %>% nrow()
run_diffs_list[[length(run_diffs_list) + 1]] <- data.frame(run1, run2, n_diff)
}
}
run_diffs <- do.call(rbind, run_diffs_list)
run_diffs
run_diffs %>% ggplot(aes(x = run1, y = run2, fill = n_diff)) +
geom_tile() + scale_fill_distiller(palette = "Spectral")
```
# Testing the less-than hypothesis
```{r}
ddsNoPrior <- nbinomWaldTest(dds, betaPrior = FALSE)
converged <- mcols(ddsNoPrior)$betaConv
converged[is.na(converged)] <- TRUE
if(any(!converged)) {
cat(sum(!converged), "non converged rows - removing\n")
print(counts_all_matrix[!converged,])
ddsNoPrior <- ddsNoPrior[converged,]
ddsNoPrior <- nbinomWaldTest(ddsNoPrior, betaPrior = FALSE)
}
res_helD_lessAbs <- results(ddsNoPrior, name = "helD_KO_lessAbs",
contrast = c("genotype","helD_KO","WT"),
lfcThreshold = 2,
altHypothesis = "lessAbs")
res_helD_lessAbs %>% as.data.frame() %>% rownames_to_column("gene") %>% filter(padj > 0.05)
plotMA(res_helD_lessAbs)
```
# Computing environment
```{r}
sessionInfo()
```