-
Notifications
You must be signed in to change notification settings - Fork 118
/
rsem-plot-model
executable file
·169 lines (141 loc) · 5.67 KB
/
rsem-plot-model
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
#!/usr/bin/env Rscript
argv <- commandArgs(TRUE)
if (length(argv) != 2) {
cat("Usage: rsem-plot-model sample_name output_plot_file\n")
q(status = 1)
}
strvec <- strsplit(argv[1], split = "/")[[1]]
token <- strvec[length(strvec)]
stat.dir <- paste(argv[1], ".stat", sep = "")
if (!file.exists(stat.dir)) {
cat("Error: directory does not exist: ", stat.dir, "\n", sep = "")
q(status = 1)
}
modelF <- paste(stat.dir, "/", token, ".model", sep = "")
cntF <- paste(stat.dir, "/", token, ".cnt", sep = "")
pdf(argv[2])
con <- file(modelF, open = "r")
# model type and forward probability
model_type <- as.numeric(readLines(con, n = 4)[1])
# fragment length distribution
strvec <- readLines(con, n = 3)
vec <- as.numeric(strsplit(strvec[1], split = " ")[[1]])
maxL <- vec[2] # maxL used for Profile
x <- (vec[1] + 1) : vec[2]
y <- as.numeric(strsplit(strvec[2], split = " ")[[1]])
mode_len = which(y == max(y)) + vec[1]
mean <- weighted.mean(x, y)
std <- sqrt(weighted.mean((x - mean)^2, y))
plot(x, y, type = "h",
main = "Fragment Length Distribution",
sub = sprintf("Mode = %d, Mean = %.1f, and Std = %.1f", mode_len, mean, std),
xlab = "Fragment Length",
ylab = "Probability")
abline(v = mode_len, col = "red", lty = "dashed")
# mate length distribution
if (model_type == 0 || model_type == 1) bval <- as.numeric(readLines(con, n = 1)[1]) else bval <- 1
if (bval == 1) {
list <- strsplit(readLines(con, n = 2), split = " ")
vec <- as.numeric(list[[1]])
maxL <- vec[2]
x <- (vec[1] + 1) : vec[2]
y <- as.numeric(list[[2]])
mode_len = which(y == max(y)) + vec[1]
mean <- weighted.mean(x, y)
std <- sqrt(weighted.mean((x - mean)^2, y))
plot(x, y, type = "h",
main = "Read Length Distribution",
sub = sprintf("Mode = %d, Mean = %.1f, and Std = %.1f", mode_len, mean, std),
xlab = "Read Length",
ylab = "Probability")
}
strvec <- readLines(con, n = 1)
# RSPD
bval <- as.numeric(readLines(con, n = 1)[1])
if (bval == 1) {
bin_size <- as.numeric(readLines(con, n = 1)[1])
y <- as.numeric(strsplit(readLines(con, n = 1), split = " ")[[1]])
par(cex.axis = 0.7)
barplot(y, space = 0, names.arg = 1:bin_size, main = "Read Start Position Distribution", xlab = "Bin #", ylab = "Probability")
par(cex.axis = 1.0)
}
strvec <- readLines(con, n = 1)
# plot sequencing errors
if (model_type == 1 || model_type == 3) {
# skip QD
N <- as.numeric(readLines(con, n = 1)[1])
readLines(con, n = N + 1)
readLines(con, n = 1) # for the blank line
# QProfile
readLines(con, n = 1)
x <- c()
peA <- c() # probability of sequencing error given reference base is A
peC <- c()
peG <- c()
peT <- c()
for (i in 1 : N) {
strvec <- readLines(con, n = 6)
list <- strsplit(strvec[1:4], split = " ")
vecA <- as.numeric(list[[1]])
vecC <- as.numeric(list[[2]])
vecG <- as.numeric(list[[3]])
vecT <- as.numeric(list[[4]])
if (sum(c(vecA, vecC, vecG, vecT)) < 1e-8) next
x <- c(x, (i - 1))
peA <- c(peA, ifelse(sum(vecA) < 1e-8, NA, -10 * log10(1.0 - vecA[1])))
peC <- c(peC, ifelse(sum(vecC) < 1e-8, NA, -10 * log10(1.0 - vecC[2])))
peG <- c(peG, ifelse(sum(vecG) < 1e-8, NA, -10 * log10(1.0 - vecG[3])))
peT <- c(peT, ifelse(sum(vecT) < 1e-8, NA, -10 * log10(1.0 - vecT[4])))
}
matplot(x, cbind(peA, peC, peG, peT), type = "b", lty = 1:4, pch = 0:3, col = 1:4,
main = "Observed Quality vs. Phred Quality Score",
xlab = "Phred Quality Score",
ylab = "Observed Quality")
legend("topleft", c("A", "C", "G", "T"), lty = 1:4, pch = 0:3, col = 1:4)
} else {
# Profile
readLines(con, n = 1)
x <- c()
peA <- c() # probability of sequencing error given reference base is A
peC <- c()
peG <- c()
peT <- c()
for (i in 1: maxL) {
strvec <- readLines(con, n = 6)
list <- strsplit(strvec[1:4], split = " ")
vecA <- as.numeric(list[[1]])
vecC <- as.numeric(list[[2]])
vecG <- as.numeric(list[[3]])
vecT <- as.numeric(list[[4]])
if (sum(c(vecA, vecC, vecG, vecT)) < 1e-8) next
x <- c(x, i)
peA <- c(peA, ifelse(sum(vecA) < 1e-8, NA, (1.0 - vecA[1]) * 100))
peC <- c(peC, ifelse(sum(vecC) < 1e-8, NA, (1.0 - vecC[2]) * 100))
peG <- c(peG, ifelse(sum(vecG) < 1e-8, NA, (1.0 - vecG[3]) * 100))
peT <- c(peT, ifelse(sum(vecT) < 1e-8, NA, (1.0 - vecT[4]) * 100))
}
matplot(x, cbind(peA, peC, peG, peT), type = "b", lty = 1:4, pch = 0:3, col = 1:4, main = "Position vs. Percentage Sequence Error", xlab = "Position", ylab = "Percentage of Sequencing Error")
legend("topleft", c("A", "C", "G", "T"), lty = 1:4, pch = 0:3, col = 1:4)
}
close(con)
# Alignment statistics
pair <- read.table(file = cntF, skip = 3, sep = "\t")
stat_len = dim(pair)[1]
upper_bound = pair[stat_len - 1, 1]
my_labels = append(0:upper_bound, pair[stat_len, 1])
my_heights = rep(0, upper_bound + 2)
dummy = sapply(1:(stat_len - 1), function(id) { my_heights[pair[id, 1] + 1] <<- pair[id, 2] })
my_heights[upper_bound + 2] = pair[stat_len, 2]
my_colors = c("green", "blue", rep("dimgrey", upper_bound - 1), "red")
barplot(my_heights, names.arg = my_labels,
col = my_colors, border = NA,
xlab = "Number of alignments per read",
ylab = "Number of reads",
main = "Alignment statistics")
pie_values = c(my_heights[1], my_heights[2], sum(my_heights[3:(upper_bound + 1)]), my_heights[upper_bound + 2])
pie_names = c("Unalignable", "Unique", "Multi", "Filtered")
pie_labels = sprintf("%s %.0f%%", pie_names, pie_values * 100.0 / sum(pie_values))
par(fig = c(0.4, 1, 0.35, 0.95), new = T)
pie(pie_values, labels = pie_labels, col = c("green", "blue", "dimgrey", "red"), clockwise = T, init.angle = 270, cex = 0.8)
par(fig = c(0, 1, 0, 1))
dev.off.output <- dev.off()