adds MultiAssayExperiment data generation

.Rbuildignore

@@ -6,4 +6,4 @@ vignettes/MANIFEST\.txt
 vignettes/Human_genes__GRCh38_p10_\.rda
 backup/
 GDCdata
-^.*\.csv
+^.*\.csv

DESCRIPTION

@@ -1,6 +1,6 @@
 Package: skcm.data
 Title: Gene expression and clinical data from Melanoma from TCGA.
-Version: 2017.11.20
+Version: 2018.06.20
 Authors@R: person("André", "Veríssimo", email = "[email protected]", role = c("aut", "cre"))
 Description: Contains the datasets for SKCM (Melanoma) with gene expression
     and clinical data. All was extracted from TCGA.
@@ -13,6 +13,9 @@ Suggests:
     futile.logger,
     TCGAbiolinks,
     knitr,
-    rmarkdown
+    rmarkdown,
+    Biobase,
+    SingleCellExperiment,
+    MultiAssayExperiment
 RoxygenNote: 6.0.1
 VignetteBuilder: knitr

NAMESPACE

@@ -1,5 +1,6 @@
 # Generated by roxygen2: do not edit by hand
 
+export(build.assay)
 export(getParticipantCode)
 export(getSampleTypeCode)
 export(joinRNASeqData)

R/functions.R

@@ -42,18 +42,14 @@ getSampleTypeCode <- function(fullBarcode) {
 #'
 #' @return a matrix with gene expression levels and all tissue samples
 #' @export
-#'
-#' @examples
 joinRNASeqData <- function() {
   # load tissue data
   data(fpkm.per.tissue)
 
   # iterate on all and join in a single matrix
-  arrays.in.dat <- sapply(seq(length(tissue)), function(ix){ is.array(tissue[[ix]]) })
-
   out.dat <- c()
-  for ( ix in which(arrays.in.dat)) {
-    out.dat <- cbind(out.dat, tissue[[ix]])
+  for (ix in names(fpkm.per.tissue)) {
+    out.dat <- cbind(out.dat, fpkm.per.tissue[[ix]])
   }
   return(out.dat)
 }
@@ -63,20 +59,12 @@ joinRNASeqData <- function() {
 #' This cannot be cached in the package as it takes too much space and
 #'  would be redundant with fpkm.per.tissue data
 #'
-#' @param project
-#' @param workflow.type
-#'
-#' @return
+#' @return a full gdc
 #' @export
-#'
-#' @examples
 loadGDCRnaSeq <- function() {
   data(gdc)
 
   temp.dat <- joinRNASeqData()
-  rownames(temp.dat) <- rownames(gdc$rnaseq)
-  temp.dat <- temp.dat[,colnames(gdc$rnaseq)]
-
   gdc$rnaseq@assays[[1]] <- temp.dat
 
   return(gdc)

R/load_assay.R

@@ -0,0 +1,84 @@
+#' Create MultiAssayExperiment object from data
+#'
+#' @param clinical use custom clinical (that can be pre-processed)
+#'
+#' @return a MultiAssayExperiment object
+#' @export
+#'
+#' @examples
+#' assay <- build.assay()
+#' assay[['RNASeq']]
+#' assar$vital_status
+build.assay <- function(clinical.custom = NULL, 
+                        gdc.custom = NULL, 
+                        mutation.custom = NULL,
+                        rnaseq.custom = NULL) {
+  # get clinical data
+  if (is.null(clinical.custom)) {
+    data(clinical)
+    clin <- clinical$all  
+  } else {
+    clin <- clinical.custom
+  }
+
+  futile.logger::flog.info('Loading \'Biospecimen\' data...')
+  if (is.null(gdc.custom)) {
+    data(gdc)
+    gdc.custom <- gdc
+  }
+
+  # get all RNASeq data
+  futile.logger::flog.info('Joining \'RNASeq\' data...')
+  if (is.null(rnaseq.custom)) {
+    rnaseq.custom <- joinRNASeqData()
+  }
+
+  futile.logger::flog.info('Loading \'Mutation\' data...')
+  if (is.null(mutation.custom)) {
+    data(mutation)
+    mutation.custom <- mutation$count
+  }
+
+  #
+  # Expression data
+
+  # map expression data with clinical
+  es.map <- data.frame(master = strtrim(colnames(rnaseq.custom), 12), 
+                       assay = colnames(rnaseq.custom), 
+                       stringsAsFactors = FALSE)
+
+  # filter only valid date.. i.e expression that have clinical data
+  valid.ix <- es.map$master %in%  clin$bcr_patient_barcode
+  valid.dat <- rnaseq.custom[, valid.ix]
+
+  sample.barcode <- strtrim(colnames(valid.dat), 16)
+  valid.codes <- sample.barcode[sample.barcode %in% gdc$bio.sample$bcr_sample_barcode]
+
+  temp.df <- Biobase::AnnotatedDataFrame(gdc$bio.sample[valid.codes,])
+  rownames(temp.df) <- colnames(valid.dat)
+
+  # build expression set
+  es  <- Biobase::ExpressionSet(assayData = valid.dat, phenoData = temp.df)
+
+  #
+  # Mutation data
+  mutation.colnames <- colnames(mutation.custom)  
+  valid.ix <- colnames(mutation.custom) %in% clin$bcr_patient_barcode
+
+  mut.map <- data.frame(master = mutation.colnames[valid.ix], assay = mutation.colnames[valid.ix])
+
+  mut <- SingleCellExperiment::SingleCellExperiment(assays = list(counts = mutation.custom))
+
+  #
+  # Setup to create MultiAssayExperiment object
+
+  futile.logger::flog.info('Building Assay...')
+  listmap <- list(es.map, mut.map)
+  names(listmap) <- c("RNASeq", "Mutation")
+
+  dfmap <- MultiAssayExperiment::listToMap(listmap)
+  objlist <- list("RNASeq" = es, "Mutation" = mut)
+  my.assay <- MultiAssayExperiment::MultiAssayExperiment(objlist, clin, dfmap)
+
+  return(my.assay)
+}

README.md

@@ -0,0 +1,70 @@
+TCGA.DATA R Package
+================
+
+-   [Package information](#package-information)
+    -   [How to use the dataset](#how-to-use-the-dataset)
+-   [How to build own data package](#how-to-build-own-data-package)
+-   [Ackowledgements](#ackowledgements)
+
+This R Package allows to retrieve Gene Expression, Mutation and clinical data from [TCGA database](http://gdc-portal.nci.nih.gov/) (The Cancer Genome Atlas). It retrieves a single type of cancer at a time.
+
+We publish diferent package in the [releases page](https://github.com/averissimo/tcga.data/releases) that allow to quickly use the datasets.
+
+The genome expression datasets are already in a matrix format ready to be used. The data is in FPKM (Fragments Per Kilobase Million) format. Any additional normalization to use in models must be performed
+
+Package information
+-------------------
+
+### How to use the dataset
+
+1.  Install `brca.data` by using `devtools` package. (`brca.data`, `prad.data` or `skcm.data`)
+
+2.  Load the library
+
+3.  Load the required datasets (one or more of the following)
+    -   `clinical`
+    -   `fpkm.per.tissue`
+    -   `fpkm.per.tissue.barcode`
+    -   `mutation`
+    -   `gdc`
+
+#### Example for BRCA package
+
+``` r
+# install the devtooks library
+install.packages('devtools')
+# The library can also be loaded and use the function install_git without 'devtools::' prefix
+devtools::install_url('https://github.com/averissimo/tcga.data/releases/download/2016.12.15-brca/brca.data_1.0.tar.gz')
+#
+# Load the brca.data package
+library(brca.data)
+# start using the data, for example the tissue data
+data(fpkm.per.tissue)
+# tissue is now in the enviromnet and will be loaded on the first
+#  time it is used. For example:
+names(fpkm.per.tissue)
+```
+
+How to build own data package
+-----------------------------
+
+1.  Open vignettes/build\_data.Rmd
+2.  Change in the header of the Rmd *(beginning of the document)* the project param to the target TCGA project
+3.  Open DESCRITION and change the name of the package to the desired name
+
+-   we use a convention of \#\#\#\#.data where \#\#\#\# is the tcga project name in lowercase
+
+1.  Run the vignettes/build\_data.Rmd to build the cache of the data
+2.  Run `devtools::document()` to create documentation
+3.  Run `devtools::build()` to build the actual package
+
+Ackowledgements
+---------------
+
+This package was developed primarily by *[André Veríssimo](http://web.tecnico.ulisboa.pt/andre.verissimo/)* with support from *Marta Lopes* and *[Susana Vinga](http://web.tecnico.ulisboa.pt/susanavinga/)*
+
+This work was supported by:
+
+-   [FCT](www.fct.pt), through IDMEC, under LAETA, projects *(UID/EMS/50022/2013)*;
+-   Susana Vinga acknowledges support by program Investigador FCT *(IF/00653/2012)* from [FCT](www.fct.pt), co-funded by the European Social Fund *(ESF)* through the Operational Program Human Potential *(POPH)*;
+-   André Veríssimo acknowledges support from [FCT](www.fct.pt) *(SFRH/BD/97415/2013)*.

vignettes/build_data.Rmd

@@ -376,14 +376,27 @@ ggplot(data = mutations.by.case) +
   scale_y_continuous(breaks = c(1, 10, 100, 1000, 10000), trans = 'log10')
 ```
 
+## MultiAssayExperiment
+
+Builds a multiAssayExperiment with clinical, rnaseq and mutation count data
+
+```{r, eval=FALSE}
+source('../R/load_assay.R')
+multiAssay <- build.assay(clinical.custom = clinical$all, 
+                          gdc.custom = gdc, 
+                          mutation.custom = mutation$count,
+                          rnaseq.custom = fpkm.per.tissue$all)
+```
+
+
 ## Exported data
 
 - `fpkm.per.tissue`: gene expression data for all types of tissues, see `names(tissue)`;
 - `fpkm.per.tissue.barcode`: Patient's participation data code (TCGA-XX-XXXX) per type of tissue;
 - `clinical`: Clinical data per tissue type. Has the same structure as tissue.
 - `mutation`: Mutation data from GDC (filtered) and a matrix of counts.
 
-```{r export_data, include=FALSE}
+```{r export_data, include=FALSE, eval=FALSE}
 # Extract all expression data to a different variable to remove redundancy in `tissue`
 fpkm.per.tissue$all <- NULL
 gdc$rnaseq     <- 'Use function loadGDCRnaSeq to get the original assay'
@@ -397,6 +410,7 @@ devtools::use_data(fpkm.per.tissue.barcode, overwrite = TRUE)
 devtools::use_data(clinical,                overwrite = TRUE)
 devtools::use_data(gene.ranges,             overwrite = TRUE)
 devtools::use_data(mutation,                overwrite = TRUE)
+#devtools::use_data(multiAssay,              overwrite = TRUE)
 ```
 
 ```{r, include=FALSE, eval=FALSE}