vignettes/prepareData.R

In SiYangming/pRRophetic: Predict clinical chemotherapeutic response from before-treatment tumor gene expression levels.

# This file explains how to acquire and prepare the data that is used in the main vignette.

library("GEOquery")

# setwd(/path/to/this/directory/)

## acquire and prepare the bortezomib data

bortezomib_mas5 <- getGEO("GSE9782")

bortIndex <- pData(phenoData(bortezomib_mas5[[1]]))[,"characteristics_ch1.1"] == "treatment = PS341"
dexIndex <- pData(phenoData(bortezomib_mas5[[1]]))[,"characteristics_ch1.1"] == "treatment = Dex"
bortData <- bortezomib_mas5[[1]]
studyIndex <- as.character(pData(bortData)[, "characteristics_ch1"])
studyResponse <- as.character(pData(bortData)[, "characteristics_ch1.8"])
exprDataBortezomib <- exprs(bortData)

# map the Ids to gene symbols
library("hgu133a.db") # version 2.8.0
x <- hgu133aSYMBOL
mapped_probes <- mappedkeys(x)
names(mapped_probes) <- as.character(x[mapped_probes])
affy2sym <- as.character(x[mapped_probes])
sym2affy <- affy2sym
names(sym2affy) <- names(affy2sym)
rownames(exprDataBortezomib) <- sym2affy[rownames(exprDataBortezomib)]


# encode CR, MR, PR, NC and PD as 1,2,3,4,5. This used in predicting from clinical data.
detailedResponse <- as.character(pData(bortData)[, "characteristics_ch1.7"])
detailedResponse[detailedResponse == "PGx_Response = IE"] <- 0
detailedResponse[detailedResponse == "PGx_Response = PD"] <- 1
detailedResponse[detailedResponse == "PGx_Response = NC"] <- 2
detailedResponse[detailedResponse == "PGx_Response = MR"] <- 3
detailedResponse[detailedResponse == "PGx_Response = PR"] <- 4
detailedResponse[detailedResponse == "PGx_Response = CR"] <- 5
detailedResponse <- as.numeric(detailedResponse)

# save(exprDataBortezomib, bortIndex, studyResponse, studyIndex, detailedResponse, file="../data/bortezomibData.RData")

## acquire and prepare the CGP data (I already have a vignette for the CGP data...)
# The drug sensitivity file is downloaded from the CGP website at: ftp://ftp.sanger.ac.uk/pub4/cancerrxgene/current_release/gdsc_manova_output_w2.csv
drugSensitivityDataCgp <- read.csv("../data/gdsc_manova_input_w2.csv", as.is=TRUE)[-1, ]

# The CEL file annotation file can be downloaded from ArrayExpress at: https://www.ebi.ac.uk/arrayexpress/files/E-MTAB-783/E-MTAB-783.sdrf.txt 
drugToCellLineDataCgp <- read.delim("../data/E-MTAB-783.sdrf.txt", as.is=TRUE)

# save(drugSensitivityDataCgp, drugSensitivityDataCgp, file="../data/cgpPhenoData.RData")

## prepare CCLE data
# The CCLE data were downloaded from the CCLE website as Gene-centric RMA-normalized mRNA expression data using BrainArray ENTREZG v15 CDF file (filename CCLE_Expression_Entrez_2012-09-29.gct). CCLE drug data was also obtained from the CCLE website (filename CCLE_NP24.2009_Drug_data_2012.02.20.csv).

# For details on how to prepare the CGP data, please see the vignette that accompanies this package. Note that to manually download the preprocess the CGP expression data, a machine with at least 32Gb of RAM will be required.