-
Notifications
You must be signed in to change notification settings - Fork 42
/
correlations_one_vs_one.Rmd
194 lines (179 loc) · 8.54 KB
/
correlations_one_vs_one.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
---
title: "Correlations between two genes across all cancers"
author: "Mikhail Dozmorov"
date: "`r Sys.Date()`"
output:
pdf_document:
toc: yes
html_document:
theme: united
toc: yes
always_allow_html: yes
editor_options:
chunk_output_type: console
---
```{r setup, echo=FALSE, message=FALSE, warning=FALSE}
# Set up the environment
library(knitr)
opts_chunk$set(cache.path='cache/', fig.path='img/', cache=F, tidy=T, fig.keep='high', echo=F, dpi=100, warnings=F, message=F, comment=NA, warning=F, results='as.is', fig.width = 10, fig.height = 6) #out.width=700,
library(pander)
panderOptions('table.split.table', Inf)
set.seed(1)
library(dplyr)
options(stringsAsFactors = FALSE)
```
```{r libraries, include=FALSE}
library(openxlsx)
library(MDmisc)
# library(org.Hs.eg.db)
# library(KEGG.db)
# library(TCGA2STAT)
library(ggplot2)
library(reshape2)
library(writexl)
# library(clusterProfiler)
# library(pathview)
# library(annotables)
# Remove non-canonical chromosome names
# grch38 <- grch38[ !(grepl("_", grch38$chr) | grepl("GL", grch38$chr)), ]
# grch38 <- grch38[, c("symbol", "description")]
# grch38 <- grch38[ complete.cases(grch38) , ]
# grch38 <- grch38[ !duplicated(grch38), ]
```
```{r functions}
# A function to load TCGA data, from remote repository, or a local R object
load_data <- function(disease = cancer, data.type = data.type, type = type, data_dir = data_dir, force_reload = FALSE) {
FILE = paste0(data_dir, "/mtx_", disease, "_", data.type, "_", type, ".rda") # R object with data
if (all(file.exists(FILE), !(force_reload))) {
# If the data has been previously saved, load it
load(file = FILE)
} else {
# If no saved data exists, get it from the remote source
mtx <- getTCGA(disease = disease, data.type = data.type, type = type, clinical = TRUE)
save(file = FILE, list = c("mtx")) # Save it
}
return(mtx)
}
# A wrapper function to perform all functional enrichment analyses.
# Helper function to save non-empty results
save_res <- function(res, fileName = fileName, wb = wb, sheetName = "KEGG") {
if (nrow(res) > 0) {
openxlsx::addWorksheet(wb = wb, sheetName = sheetName)
openxlsx::writeData(wb, res, sheet = sheetName)
openxlsx::saveWorkbook(wb, fileName, overwrite = TRUE)
}
}
# A wrapper to save the results
save_enrichr <- function(up.genes = up.genes, dn.genes = NULL, databases = "KEGG_2016", fdr.cutoff = 1, fileNameOut = NULL, wb = NULL) {
print(paste("Running", databases, "analysis", sep = " "))
if (is.null(dn.genes)) {
res.kegg <- enrichGeneList(up.genes, databases = databases, fdr.cutoff = 1)
} else {
res.kegg <- enrichFullGeneList(up.genes, dn.genes, databases = databases, fdr.cutoff = 1)
}
res.kegg$pval <- formatC(res.kegg$pval, digits = 3, format = "e")
res.kegg$qval <- formatC(res.kegg$qval, digits = 3, format = "e")
if (!is.null(fileNameOut)) {
if (nchar(databases) > 30) databases <- paste0(substr(databases, 1, 20), "_", substr(databases, nchar(databases) - 8, nchar(databases))) # If a database is longer that 30 characters, keep first 20 and last 10 characters
save_res(res.kegg, fileNameOut, wb = wb, sheetName = databases)
}
# Pause for a few seconds
pause_sec <- round(runif(1, min = 1, max = 10))
Sys.sleep(pause_sec)
return(res.kegg)
}
```
```{r settings}
system("mkdir -p data")
system("mkdir -p results")
# Path where the downloaded data is stored
data_dir = "/Users/mdozmorov/Documents/Data/GenomeRunner/TCGAsurvival/data" # Mac
# data_dir = "F:/Data/GenomeRunner/TCGAsurvival/data" # Windows
# Selected genes, TWO genes only
selected_genes <- c("IGFBP3", "TMEM219")
# Check if the names exist
genes <- readLines("data.TCGA/TCGA_genes.txt")
setdiff(selected_genes, genes) # Should be 0, meaning selected genes are among the TCGA genes
data.type = "RNASeq2"; type = ""
# All cancers with RNASeq2 data
cancer_RNASeq2 = c("ACC", "BLCA", "BRCA" , "CESC", "CHOL", "COAD", "COADREAD", "DLBC", "ESCA", "GBM", "GBMLGG", "HNSC", "KICH", "KIPAN", "KIRC", "KIRP", "LGG", "LIHC", "LUAD", "LUSC", "MESO", "OV", "PAAD", "PCPG", "PRAD", "READ", "SARC", "SKCM", "STAD", "TGCT", "THCA", "THYM", "UCEC", "UCS")
# fileNameIn <- (paste0("data/All_expression_", data.type, "_", type, ".Rda")) # Save expression data
# fileNameOut <- paste0("results/All_correlation_", selected_genes, "_", data.type, "_", type, ".Rda") # Save correlation data
# fileNameRes <- paste0("results/All_results_", selected_genes, "_", data.type, "_", type, ".xlsx") # Save results
# Or, one cancer
# cancer_RNASeq2 = c("LUAD")
fileNameIn <- (paste0("data/", cancer_RNASeq2, "_expression_", data.type, "_", type, ".Rda")) # Save expression data
fileNameOut <- paste0("results/correlations_", selected_genes[1], "_vs_", selected_genes[2], ".xlsx") # Save correlation results
# Correlation type
corr_type <- "pearson"
# Correlation cutoffs
corr_cutoff <- 0.2
pval_cutoff <- 0.05
# Enrichment cutoffs
p.adj.cutoff <- 0.1
fdr.cutoff <- 0.3
```
- Each cancer is processed separately. Headers "Processing cancer XXX" start results section for a given cancer.
- Linear regression results between expression of two genes are outputted. "Adj R2" and "P" - percent variability explained by regression model (higher the better) and its p-value (smaller the better).
- Correlation graph shows expression of one gene plotted vs. the other gene in a given cancer, with linear regression line and confidence intervals around it.
- Expression graph shows boxplots of expression of two genes in a given cancer. If one, or both, genes are low expressed, any correlation/regression results should be interpreted with caution.
```{r correlations, fig.height=3, fig.width=3}
all_corrs <- list() # List to store cancer-specific correlationa
all_pvals <- list() # List to store cancer-specific p-values
# Get correlation matrixes for the gene of interest in each cancer
for (cancer_type in cancer_RNASeq2) {
pandoc.header(paste0("Processing cancer ", cancer_type))
# Prepare expression data
mtx <- load_data(disease = cancer_type, data.type = data.type, type = type, data_dir = data_dir, force_reload = FALSE)
expr <- mtx$merged.dat[ , 4:ncol(mtx$merged.dat)] %>% as.matrix
# Attach column with gene symbols
expr <- data.frame(hgnc = colnames(expr), t(expr))
# Correlation for a pair of genes
cancer_rcorr <- Hmisc::rcorr(as.numeric(expr[selected_genes[1], 2:ncol(expr)]),
as.numeric(expr[selected_genes[2], 2:ncol(expr)]),
type = corr_type)
# Matrix to plot correlations
mtx_to_plot <- data.frame(gene1 = log2(as.numeric(expr[selected_genes[1], 2:ncol(expr)]) + 1),
gene2 = log2(as.numeric(expr[selected_genes[2], 2:ncol(expr)]) + 1))
colnames(mtx_to_plot) <- c(selected_genes[1], selected_genes[2]) # Assign column names
# Quick regression
fit <- lm(mtx_to_plot[, 1] ~ mtx_to_plot[, 2])
# Plot correlation graph
p <- ggplot(mtx_to_plot, aes(x = eval(parse(text = selected_genes[1])), y = eval(parse(text = selected_genes[2])))) +
geom_point(shape = 1) +
geom_smooth(method = lm) +
xlab(selected_genes[1]) +
ylab(selected_genes[2]) +
ggtitle("Correlation graph")
# Fit statistics
print(paste("Adj R2 = ", signif(summary(fit)$adj.r.squared, 5),
"Intercept =", signif(fit$coef[[1]],5 ),
" Slope =", signif(fit$coef[[2]], 5),
" P =", signif(summary(fit)$coef[2,4], 5)))
plot(p)
# Plot expression graph
mtx_to_plot <- melt(mtx_to_plot)
p <- ggplot(mtx_to_plot, aes(x = variable, y = value, fill = variable)) +
geom_boxplot() +
ggtitle("Expression plot") +
xlab("Gene") + ylab("log2 expression")
plot(p)
# Store the results
all_corrs <- c(all_corrs, list(cancer_rcorr$r[1, 2]))
all_pvals <- c(all_pvals, list(cancer_rcorr$P[1, 2]))
names(all_corrs)[length(all_corrs)] <- names(all_pvals)[length(all_pvals)] <- cancer_type
}
```
# Correlation (`r corr_type`) between `r selected_genes[1]` and `r selected_genes[2]`
- "corr", "pval" - `r corr_type` correlation coefficient and the p-value of correlations
- Sort the table by "corr", "pval" columns
```{r}
res_corr <- data.frame(Acronym = cancer_RNASeq2, corr = unlist(all_corrs), pval = unlist(all_pvals))
cancers <- openxlsx::read.xlsx("data.TCGA/TCGA_cancers.xlsx")
res_corr <- left_join(res_corr, cancers, by = c("Acronym"))
res_corr <- round_df(res_corr)
res_corr <- res_corr[ order(res_corr$corr, decreasing = TRUE), ]
# DT::datatable(res_corr, options = list(pageLength = 50))
pander(res_corr)
write_xlsx(list(Correlation = res_corr), fileNameOut)
```