Newer
Older
#######################################################
##### seed eQTL network analysis using ARACNe #########
#######################################################
# prepare the script working environment ####
remove(list = ls())
gc()
set.seed(1000)
work.dir <- "C:/Users/harta005/Projects/seed-germination-qtl"
setwd(work.dir)
# library
library(RCy3)
library(dplyr)
library(minet)
library(GENIE3)
library(doParallel)
library(reshape2)
library(devtools)
library(ggplot2)
library(Hmisc)
# determine the stage and transband ####
Hartanto, Margi
committed
dev.stage <- 'rp'
Hartanto, Margi
committed
start <- 24000000
end <- 26000000
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
n.perm <- 1000
# load required data ####
# gene and genetic infp
genetic.map <- as.matrix(read.csv(file = 'files/genetic-map.csv'))
gene.info <- read.csv('files/gene.info.csv', row.names = 1)
gene.feature <- read.csv('files/gene-feature.csv')
gene.pol <- read.csv('files/baysha-polymorphism.csv', row.names = 1)
gene.data <- merge(gene.info, gene.pol, all = TRUE)
gene.feature <- merge(gene.info, gene.feature, all = TRUE)
# expression matrix
expression <- as.matrix(read.csv('files/trait-matrix.csv', row.names = 1))
expression <- expression[, colnames(expression) %in% colnames(genetic.map)] # remove sample without genetic map
expression <- expression[, 17:176] # remove parents
expression <- expression[, grep(pattern = 'pd', x = colnames(expression))] # only take ril for specific stage
# ril stage
ril.stage <- substr(x = colnames(expression), start = 8, stop = 10) # write a vector of ril stage
# load the qtl table
eqtl.table <- readRDS('qtl-tables/table_single-stage-eqtl_all.rds')
# select the target trait
target.table <- filter(eqtl.table, qtl_bp >= start,
qtl_type == 'distant',
qtl_bp <= end,
qtl_chromosome == chr,
stage == dev.stage)
targets <- as.character(target.table$trait)
target.expression <- expression[which(rownames(expression) %in% targets), ]
# select the candidate genes
candidate.table <- filter(eqtl.table, qtl_bp >= start,
qtl_bp <= end,
qtl_type == 'local',
qtl_chromosome == chr,
gene_bp >= start,
gene_bp <= end,
stage == dev.stage)
candidates <- as.character(candidate.table$trait)
candidate.expression <- expression[which(rownames(expression) %in% rownames(expression)), ]
# combine expression matrix for targets, candidates, and genes underlying qtl
expression.m <- expression[which(rownames(expression) %in% c(targets, candidates)), ]
# create gene table and determine the gene function
nodes <- gene.feature[which(gene.feature$trait %in% c(targets, candidates)), ]
nodes <- data.frame(id = c(targets, candidates),
role = c(rep('target', length(targets)), rep('candidates', length(candidates))),
is_tf = NA)
nodes[, 'is_tf'] <- gene.feature[match(nodes$id, gene.feature$trait), 'is_TF']
regulators <- as.character(nodes$trait[which(nodes$role == 'candidates' | nodes$is_tf == 1)])
# network inference using GENIE3 ####
weight <- GENIE3(exprMatrix = expression.m, verbose = TRUE)
weight <- melt(weight)
weight$Var1 <- as.character(weight$Var1)
weight$Var2 <- as.character(weight$Var2)
names(weight) <- c('nodes1', 'nodes2', 'weight')
# remove duplicates
edges.genie3 <- as.data.frame(matrix(data = NA, nrow = 6670, ncol = 3))
cur.row <- 1
for(i in row.names(expression.m)) {
for(j in row.names(expression.m)) {
if(i == j) {
next
}
set1 <- weight[which(weight$nodes1 == i & weight$nodes2 == j), ]
set2 <- weight[which(weight$nodes2 == i & weight$nodes1 == j), ]
if(set1[, 3] >= set2[, 3]) {
edges.genie3[cur.row, ] <- set1
cur.row <- cur.row + 1
print(cur.row)
}
if(set1[, 3] <= set2[, 3]) {
edges.genie3[cur.row, ] <- set2
cur.row <- cur.row + 1
print(cur.row)
}
}
}
edges.genie3 <- distinct(edges.genie3)
names(edges.genie3) <- c('nodes1', 'nodes2', 'weight')
# ranking
rank.genie3 <- edges.genie3
rank.genie3$rank <- rank(-rank.genie3$weight)
rank.genie3 <- rank.genie3[, c(1, 2, 4)]
write.csv(x = rank.genie3, file = paste0('networks/', dev.stage, chr, '-genie.csv'))
# network inference using pearson correlation ####
edges.pearson <- rcorr(t(expression.m), type = 'pearson')
edges.pearson <- melt(edges.pearson$r)
names(edges.pearson) <- c('nodes1', 'nodes2', 'weight')
# remove duplicates and determine direction
rank.pearson <- rank.genie3[, 1:2]
rank.pearson <- merge(rank.pearson, edges.pearson, by = c("nodes1", "nodes2"))
rank.pearson <- rank.pearson[, 1:3]
rank.pearson$weight <- abs(rank.pearson$weight)
# ranking
rank.pearson$rank <- rank(-rank.pearson$weight)
rank.pearson <- rank.pearson[, c(1, 2, 4)]
names(rank.pearson) <- c('nodes1', 'nodes2', 'rank')
write.csv(x = rank.pearson, file = paste0('networks/', dev.stage, chr, '-pearson.csv'))
# network inference using spearman correlation ####
edges.spearman <- rcorr(t(expression.m), type = 'spearman')
edges.spearman <- melt(edges.spearman$r)
names(edges.spearman) <- c('nodes1', 'nodes2', 'weight')
# remove duplicates and determine direction
rank.spearman <- rank.genie3[, 1:2]
rank.spearman <- merge(rank.spearman, edges.spearman, by = c("nodes1", "nodes2"))
rank.spearman <- rank.spearman[, 1:3]
rank.spearman$weight <- abs(rank.spearman$weight)
# ranking
rank.spearman$rank <- rank(-rank.spearman$weight)
rank.spearman <- rank.spearman[, c(1, 2, 4)]
names(rank.spearman) <- c('nodes1', 'nodes2', 'rank')
write.csv(x = rank.spearman, file = paste0('networks/', dev.stage, chr, '-spearman.csv'))
# network inference using CLR ####
edges.clr <- minet(dataset = t(expression.m), method = 'clr', estimator = 'mi.empirical', disc = 'equalfreq')
edges.clr <- melt(edges.clr)
names(edges.clr) <- c('nodes1', 'nodes2', 'weight')
# ranking
rank.clr <- rank.genie3[, 1:2]
rank.clr <- merge(rank.clr, edges.clr, by = c("nodes1", "nodes2"))
rank.clr <- rank.clr[, 1:3]
rank.clr$rank <- rank(-rank.clr$weight)
rank.clr <- rank.clr[, c(1, 2, 4)]
names(rank.clr) <- c('nodes1', 'nodes2', 'rank')
write.csv(x = rank.clr, file = paste0('networks/', dev.stage, chr, '-clr.csv'))
# network inference using aracne ####
edges.aracne <- minet(dataset = t(expression.m), method = 'aracne', estimator = 'mi.empirical', disc = 'equalfreq')
edges.aracne <- melt(edges.aracne)
names(edges.aracne) <- c('nodes1', 'nodes2', 'weight')
# ranking
rank.aracne <- rank.genie3[, 1:2]
rank.aracne <- merge(rank.aracne, edges.aracne, by = c("nodes1", "nodes2"))
rank.aracne <- rank.aracne[, 1:3]
rank.aracne$rank <- rank(-rank.aracne$weight)
rank.aracne <- rank.aracne[, c(1, 2, 4)]
names(rank.aracne) <- c('nodes1', 'nodes2', 'rank')
write.csv(x = rank.aracne, file = paste0('networks/', dev.stage, chr, '-aracne.csv'))
# network inference using mrnet ####
edges.mrnet <- minet(dataset = t(expression.m), method = 'mrnet', estimator = 'mi.empirical', disc = 'equalfreq')
edges.mrnet <- melt(edges.mrnet)
names(edges.mrnet) <- c('nodes1', 'nodes2', 'weight')
# ranking
rank.mrnet <- rank.genie3[, 1:2]
rank.mrnet <- merge(rank.mrnet, edges.mrnet, by = c("nodes1", "nodes2"))
rank.mrnet <- rank.mrnet[, 1:3]
rank.mrnet$rank <- rank(-rank.mrnet$weight)
rank.mrnet <- rank.mrnet[, c(1, 2, 4)]
names(rank.mrnet) <- c('nodes1', 'nodes2', 'rank')
write.csv(x = rank.mrnet, file = paste0('networks/', dev.stage, chr, '-mrnet.csv'))
# network inference using tigress
weight.tigress <- tigress::tigress(t(expression.m),
nsplit = 1000,
nstepsLARS = 5,
allsteps = F)
weight.tigress <- melt(weight.tigress)
names(weight.tigress) <- c('nodes1', 'nodes2', 'weight')
weight.tigress$nodes1 <- as.character(weight.tigress$nodes1)
weight.tigress$nodes2 <- as.character(weight.tigress$nodes2)
# remove duplicates
# combination <- nrow(expression.m) * (nrow(expression.m) - 1) /2
# edges.tigress <- as.data.frame(matrix(data = NA, nrow = combination, ncol = 3))
# cur.row <- 1
# for(i in row.names(expression.m)) {
# for(j in row.names(expression.m)) {
# if(i == j) {
# next
# }
# set1 <- weight.tigress[which(weight.tigress$nodes1 == i & weight.tigress$nodes2 == j), ]
# set2 <- weight.tigress[which(weight.tigress$nodes2 == i & weight.tigress$nodes1 == j), ]
# if(set1[, 3] >= set2[, 3]) {
# edges.tigress[cur.row, ] <- set1
# cur.row <- cur.row + 1
# print(cur.row)
# }
# if(set1[, 3] <= set2[, 3]) {
# edges.genie3[cur.row, ] <- set2
# cur.row <- cur.row + 1
# print(cur.row)
# }
# }
# }
# edges.tigress <- distinct(edges.tigress)
# names(edges.tigress) <- c('nodes1', 'nodes2', 'weight')
# ranking
rank.tigress <- rank.genie3[, 1:2]
rank.tigress <- merge(rank.tigress, weight.tigress, by = c("nodes1", "nodes2"))
rank.tigress <- rank.tigress[, 1:3]
rank.tigress$rank <- rank(-rank.tigress$weight)
rank.tigress <- rank.tigress[, c(1, 2, 4)]
names(rank.tigress) <- c('nodes1', 'nodes2', 'rank')
write.csv(x = rank.tigress, file = paste0('networks/', dev.stage, chr, '-tigress.csv'))
# merge the tables
rank.list <- list(rank.genie3, rank.spearman, rank.clr, rank.aracne, rank.tigress)
#rank.list <- list(rank.genie3, rank.spearman, rank.pearson, rank.mrnet, rank.clr, rank.aracne, rank.tigress)
networks <- Reduce(function(x, y) full_join(x, y, by = c('nodes1', 'nodes2')), rank.list)
names(networks) <- c('nodes1', 'nodes2', 'genie3', 'spearman', 'clr', 'aracne', 'tigress')
#names(networks) <- c('nodes1', 'nodes2', 'genie3', 'spearman', 'pearson', 'mrnet', 'clr', 'aracne', 'tigress')
networks$avg_rank <- NA
for(i in 1:nrow(networks)) {
networks$avg_rank[i] <-
mean(as.numeric(networks[i, 3:(ncol(networks) - 1)]))
}
# PCA
pr.out <- prcomp(x = (t(networks[3:10])), center = T, scale. = F)
pc.df <- data.frame(pc1 = pr.out$x[, 1], # 0.37
pc2 = pr.out$x[, 2]) # 0.32
pc.df$method <- rownames(pc.df)
ggplot(pc.df, aes(x = pc1, y = pc2, color = method)) + geom_point(size = 3.5)
pc.sum <- summary(pr.out)
pc1 <- pc.sum$importance[2, 1]
pc2 <- pc.sum$importance[2, 2]
# determine threshold
edges <- networks[, c('nodes1', 'nodes2', 'avg_rank')]
colnames(edges) <- c('source', 'target', 'interaction')
edges <- na.omit(edges)
threshold.m <- as.data.frame(matrix(data = NA, ncol = 3, nrow = 0))
names(threshold.m) <- c('threshold', "edge", "node")
cur.row <- 1
for(i in max(round(edges$interaction, 0)):0){
edges.tmp <- filter(edges, interaction <= i)
total.edge <- nrow(edges.tmp)
total.node <- length(unique(c(edges.tmp$source, edges.tmp$target)))
threshold.m[cur.row, ] <- c(i, total.edge, total.node)
cur.row <- cur.row + 1
}
# visualize the threshold matrix
plot(threshold.m$threshold,
threshold.m$node,
xlab = "rank threshold",
Hartanto, Margi
committed
ylab = 'number of node') # 1552 for RP4 and 1248 for PD2
# determine the stability of rank (RP4)
degree.table <- data.frame(candidates = candidates)
for(i in max(round(edges$interaction, 0)):1) {
print(i)
edge.thresholded <- filter(edges, interaction <= i)
degree.table.tmp <- as.data.frame(table(edge.thresholded$source))
colnames(degree.table.tmp) <- c("candidates", i)
rank.degree.table <- merge(data.frame(candidates = candidates),
degree.table.tmp, all.x = T)
rank.degree.table$rank <- rank(-rank.degree.table[, 2],
na.last = T, ties.method = 'average')
degree.table[, paste0(i)] <- rank(-rank.degree.table[, 2],
na.last = T, ties.method = 'average')
}
all.rank <- data.frame(candidates,
mean = round(apply(degree.table[2:ncol(degree.table)], 1, mean), 2),
sd = round(apply(degree.table[2:ncol(degree.table)], 1, sd), 2))
all.rank <- all.rank[order(all.rank$mean), ]
write.csv(degree.table, paste0('networks/', dev.stage, chr, '-ranks-for-all-threshold.csv'))
Hartanto, Margi
committed
write.csv(all.rank, paste0('networks/', dev.stage, chr, '-ranks.csv'))
# visualize the network
edges.th <- filter(edges, interaction <= 288)
#edges$interaction <- log2(edges$interaction)
write.csv(edges, paste0('networks/', dev.stage, chr, '-edge.csv'))
write.csv(nodes, paste0('networks/', dev.stage, chr, '-node.csv'))
createNetworkFromDataFrames(nodes = nodes, edges = edges.th)