-
Notifications
You must be signed in to change notification settings - Fork 0
/
Split_script.Rmd
179 lines (147 loc) · 4.98 KB
/
Split_script.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
---
title: "Split_script"
author: "Lorena"
date: " "
output:
word_document: default
html_document: default
---
```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```
```{r}
library(tidyverse)
library(plyr)
library(dplyr)
library(caret)
```
```{r}
#prepare directory
setwd(".")
ROOT <- "."
# TCGA data directory
DATADIR <- file.path(ROOT, "Data")
# ACGT data dir (for clinical info)
ACGTDIR <- file.path(ROOT, "original")
# output directory
OUTROOT <- DATADIR
```
```{r}
tumor <- "AML"
read_tcga <- function(datafile, sample_col="ID", suffix="01") {
df <- as.data.frame(read_delim(datafile, " ", progress=FALSE)) %>%
tibble::column_to_rownames(sample_col) %>%
dplyr::select(ends_with(suffix))
# variance filter
vars <- apply(df, 1, IQR)
quant <- quantile(vars, probs=0.5)
selected <- !is.na(vars) & vars > quant
as.data.frame(t(df[selected, ])) %>% tibble::rownames_to_column(var="Sample")
}
```
```{r}
# list of ID suffixes to keep (tumor-dependent)
suffixes <- list()
suffixes[["AML"]] <- "03"
samples <- read.delim2("./original/clinical/aml", header = TRUE, sep = "\t")
expr <- read_tcga(file.path(ACGTDIR, str_to_lower(tumor), "exp"), suffix=suffixes[[tumor]])
meth <- read_tcga(file.path(ACGTDIR, str_to_lower(tumor), "methy"), suffix=suffixes[[tumor]])
mirna <- read_tcga(file.path(ACGTDIR, str_to_lower(tumor), "mirna"), suffix=suffixes[[tumor]])
samples <- samples %>% setNames(paste0("clin:", names(.))) %>%
dplyr::rename(Sample=`clin:sampleID`)
```
```{r}
# throw away columns with all NAs
samples <- samples[, colSums(is.na(samples)) != nrow(samples)]
samples <- samples[grepl(sprintf("%s$", suffixes[[tumor]]), samples$Sample), ]
samples$Sample <- gsub("-", ".", samples$Sample)
expr <- expr %>% setNames(paste0("gene:", names(.))) %>%
dplyr::rename(Sample=`gene:Sample`)
meth <- meth %>% setNames(paste0("meth:", names(.))) %>%
dplyr::rename(Sample=`meth:Sample`)
mirna <- mirna %>% setNames(paste0("mirna:", names(.))) %>%
dplyr::rename(Sample=`mirna:Sample`)
```
```{r}
#save data
write_csv(expr, "./original/aml/expCl.txt")
write_tsv(meth, "./original/aml/methCl.txt")
write_tsv(mirna, "./original/aml/mirnaCl.txt")
```
```{r}
# get common samples
omics <- list(expr=expr$Sample, meth=meth$Sample, mirna=mirna$Sample)
sampleIntersect <- Reduce(intersect, omics)
samples <- samples[samples$Sample %in% sampleIntersect, ]
# join all dataframes
mrg <- join_all(list(samples, expr, meth, mirna), by="Sample", type="left")
# remove near-zero variance features
nzv <- nearZeroVar(mrg, saveMetrics=TRUE, foreach=TRUE)
mrg <- mrg[, !nzv$nzv]
# save
write_tsv(mrg, file.path(OUTROOT, tumor, "merged.txt"))
```
```{r}
# select
tgt <- "clin:X_OS_IND"
mrg.sub <- mrg[!is.na(mrg[, tgt]), ]
y <- factor(mrg.sub[, tgt])
mrg.sub <- dplyr::select(mrg.sub, -`tgt`)
table(y) # 0 = alive, 1 = deceased
```
```{r}
colMeans(is.na(mrg.sub))
```
```{r}
# elimino las variables clinicas con mas de un 10% de NA
porcentaje <- .1
columnas_a_borrar <- which(colMeans(is.na(mrg.sub))>porcentaje)
mrg.sub[,columnas_a_borrar] <- NULL
```
```{r}
#eliminar<- select(mrg.sub,starts_with("clin:X_GENOMIC_ID_"))
mrg.sub<- select(mrg.sub, -c(3,9, 38, 42:58))# elimino otras variables de identificación del paciente
#save
write_csv(mrg.sub, "./VAEdata/mrg_sub.csv")
```
```{r}
# elimino la variable donde indica si el pciente sobrevive o no.
allData<- select(mrg.sub, -c("clin:vital_status"))
#save
write_csv(allData, "./VAEdata/allData.csv")
```
```{r}
omicsData<-select(allData,starts_with(c("Sample","gene:","meth","mirna")))
#colnames(df) <- gsub(" ","",colnames(df))
colnames(omicsData) <- gsub(" ","",colnames(omicsData))
head(omicsData)
#save
write_csv(omicsData, "./VAEdata/omicsData.csv")
```
```{r}
for(split.id in seq(1, 10)) {
outdir <- file.path(OUTROOT,"data","splitOmics", split.id)
if(!dir.exists(outdir))
dir.create(outdir, recursive=TRUE)
# make it so that the 1st split is for set.seed(78), which was the one created previously
set.seed(78+split.id)
train.idx <- createDataPartition(y=y, p=0.7, list=FALSE)
train.data <- omicsData[train.idx,]
train.lab <- y[train.idx]
test.data <- omicsData[-train.idx,]
test.lab <- y[-train.idx]
# write labels
write.table(train.lab, file=file.path(outdir, "labels_OS_tr.txt"), sep="\t", quote=FALSE, row.names=F, col.names=F)
write.table(test.lab, file=file.path(outdir, "labels_OS_ts.txt"),sep="\t", quote=FALSE, row.names=F, col.names=F)
write_tsv (train.data, file.path(outdir, "omicsData_train.csv"))
write_tsv(test.data,file.path(outdir, "omicsData_test.csv"))
}
```
```{r}
omicsDataOS <- mrg.sub[,-(2:36),drop=FALSE]
omicsDataOS[3] <- NULL
colnames(omicsDataOS) <- gsub(" ","",colnames(omicsDataOS))
head(omicsDataOS)
#save
write_csv(omicsDataOS, "./VAEdata/omicsDataOS.csv")
```