-
Notifications
You must be signed in to change notification settings - Fork 1
/
05r-dlbcl_feature_selection.R
224 lines (173 loc) · 8.07 KB
/
05r-dlbcl_feature_selection.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
################################################################################
################################################################################
################################################################################
################################################################################
############################ DLBCL CLUSTER FEATURES ############################
#################################### SETUP #####################################
library(tidyverse)
library(DESeq2)
library(matrixStats)
library(glmnet)
library(Boruta)
library(UpSetR)
library(c060)
library(randomForest)
library(rpart)
library(rpart.plot)
################################### LOAD DATA ##################################
load("r_outputs/02-DLBCL_filt_counts.Rdata")
load("r_outputs/05o-DLBCL_pca_ccp_clusters_metadata.Rdata")
load("r_outputs/01-refs.Rdata")
################################################################################
################################################################################
#################################### CLUST 7 ###################################
################################################################################
################################################################################
################################# METADATA SETUP ###############################
DLBCL_metadata$clust.retro.k7 <- clust.df$clust.retro.k7
#################################### DESEQ2 ####################################
countDat <- DLBCL.filt.herv
cat(sprintf('%d variables\n', nrow(countDat)))
stopifnot(all(colnames(countDat) == rownames(DLBCL_metadata)))
dds <- DESeq2::DESeqDataSetFromMatrix(countDat, DLBCL_metadata, ~1 )
dds <- DESeq2::DESeq(dds, parallel=T)
tform <- DESeq2::varianceStabilizingTransformation(dds, blind=FALSE)
##################################### LRT ######################################
p.cutoff <- 1e-3
lfc.cutoff <- 1.5
dds_lrt <- DESeq2::DESeqDataSetFromMatrix(countDat,DLBCL_metadata, ~ clust.retro.k7)
dds_lrt <- DESeq2::DESeq(dds_lrt, test="LRT", reduced = ~ 1, parallel=T)
res_lrt <- DESeq2::results(dds_lrt, alpha=p.cutoff)
sig_lrt <- subset(res_lrt, padj < p.cutoff & abs(log2FoldChange) > lfc.cutoff)
selected_vars <- list()
selected_vars$lrt <- rownames(sig_lrt)
############################### FILTER VARIABLES ###############################
ntop <- nrow(tform) # No filtering
vars <- rowVars(assay(tform))
mat <- assay(tform)[order(-vars)[1:ntop], ]
cat(sprintf('%d variables\n', nrow(mat)))
#################################### BORUTA ####################################
DLBCL_metadata$clust.retro.k7 <- as.factor(DLBCL_metadata$clust.retro.k7)
set.seed(12345)
bor.orig <- Boruta(x=t(mat), y=DLBCL_metadata$clust.retro.k7, doTrace=2, ntree=1000, maxRuns=1000)
print(bor.orig)
bor.model <- TentativeRoughFix(bor.orig)
print(bor.model)
selected_vars$boruta <- names(bor.model$finalDecision)[bor.model$finalDecision == 'Confirmed']
#################################### LASSO #####################################
weakness <- 1 # For each subsample the features are reweighted by a random weight uniformly sampled in [weakness,1]
size <- 1 # proportion of samples drawn in every subsample
steps <- 100 # number of subsamples
error <- 0.01 # the desired type I error level w.r.t. to the chosen type I error rate.
pi_thr <- 0.6 # the threshold used for the stability selection
ncores <- 1 # force run on 1 core, prevents reseeding RNG on each core
set.seed(12345)
res.stabpath.c <-
c060::stabpath(
DLBCL_metadata$clust.retro.k7,
t(mat),
mc.cores = ncores,
weakness=weakness,
family = "multinomial",
type.multinomial = "grouped"
)
plot(res.stabpath.c)
res.stabsel.c <-
c060::stabsel(
res.stabpath.c,
error = 0.1,
pi_thr = 0.6,
type = "pfer"
)
length(res.stabsel.c$stable)
selected_vars$lasso <- names(res.stabsel.c$stable)
pdf("plots/05r-DLBCL_features_upset.pdf", height=5, width=7)
UpSetR::upset(fromList(selected_vars),
sets=c('lrt', 'boruta', 'lasso'),
order.by = "freq",
text.scale = c(2, 2, 2, 2, 2, 2))
dev.off()
################################# FEATURE GRAPH ################################
mat.sel <- mat[Reduce(intersect, list(selected_vars$lrt, selected_vars$boruta, selected_vars$lasso)), ]
mat.sel <- mat[selected_vars$lasso, ]
# "HERVW_2q32.3" "HML2_7p22.1" "HERVH_7q11.23a" "HERVH_16p13.2e"
rpart.fit <- rpart(y ~ . , data=data.frame(y=DLBCL_metadata$clust.retro.k7,
t(mat.sel)),
method = 'class')
pdf("plots/05r-DLBCL_features_rpart.pdf", height=7, width=9)
rpart.plot(rpart.fit, type=5, extra=1, digits=3)
dev.off()
################################# PLOT COUNTS ##################################
plot.counts <- function(df, gene, title) {
as.data.frame(plotCounts(df,
gene=gene,
intgroup="clust.retro.k7",
returnData = TRUE)) %>%
ggplot(aes(x=clust.retro.k7, y=count, fill=clust.retro.k7)) +
geom_boxplot(notch = TRUE) +
theme_pubr() +
theme(legend.position="none",
axis.text.x = element_text(angle=45, hjust=1)) +
xlab("HERV Cluster") +
ylab("Counts") +
ggtitle(title) +
theme(plot.title = element_text(hjust = 0.5),
aspect.ratio = 1) +
ylim(0,10000) +
scale_fill_manual(values = c("HC1" = "#E64B35FF",
"HC2" = "#4DBBD5FF",
"HC3" = "#00A087FF",
"HC4" = "#3C5488FF",
"HC5" = "#F39B7FFF",
"HC6" = "#8491B4FF",
"HC7" = "#91D1C2FF")) +
scale_y_log10(labels = label_comma())
}
p1 <- plot.counts(dds_lrt, "HML2_7p22.1", "HML2_7p22.1")+
stat_compare_means(comparisons = list(c("HC4", "HC6"),
c("HC4", "HC5"),
c("HC4", "HC7")),
method = "t.test",
label = "p.signif")
p2 <- plot.counts(dds_lrt, "HERVH_16p13.2e", "HERVH_16p13.2e") +
stat_compare_means(comparisons = list(c("HC7", "HC6"),
c("HC7", "HC5"),
c("HC7", "HC4"),
c("HC7", "HC3"),
c("HC7", "HC2"),
c("HC7", "HC1")),
method = "t.test",
label = "p.signif")
p3 <- plot.counts(dds_lrt, "HERVW_2q32.3", "HERVW_2q32.3") +
stat_compare_means(comparisons = list(c("HC1", "HC2"),
c("HC2", "HC3"),
c("HC2", "HC7")),
method = "t.test",
label = "p.signif")
p4 <- plot.counts(dds_lrt, "HERVH_7q11.23a", "HERVH_7q11.23a") +
stat_compare_means(comparisons = list(c("HC3", "HC5"),
c("HC5", "HC7"),
c("HC2", "HC5")),
method = "t.test",
label = "p.signif")
pdf("plots/05r-DLBCL_lasso_selected_features.pdf", height=6, width=9)
plot_grid(p1,
p2,
p3,
p4,
nrow = 2,
ncol = 2)
dev.off()
plot.counts(dds_lrt, "HARLEQUIN_1q32.1", "HARLEQUIN_1q32.1")+
stat_compare_means(comparisons = list(c("C1", "C2"),
c("C2", "C3"),
c("C2", "C4"),
c("C2", "C5"),
c("C2", "C7"),
c("C2", "C6")),
method = "t.test",
label = "p.signif")
#################################### SAVE DATA #################################
save(selected_vars, mat.sel, dds_lrt, sig_lrt, bor.model, res.stabsel.c,
file="r_outputs/05r-dlbcl_selected_features.Rdata")
load("r_outputs/05r-dlbcl_selected_features.Rdata")