Hi,

I would like to add a matrix to the DGEList in a similar way assays can be added to a SummarizedExperiment.

I calculated the log transformed cpm values of a DGEList and I would like to store them in the DGEList, so subsetting this object would subset the log transformed cpm values too.

here is my code:

query <- GDCquery(project = "TCGA-BLCA",  # project-code for Bladder Cancer
                  data.category = "Transcriptome Profiling",  # catogry for RNA-Seq Data
                  data.type = "Gene Expression Quantification", # raw reads, or , aligned data
                  workflow.type = "HTSeq - Counts") # FPKM / HTSeq-Count / etc.

GDCdownload(query, # name of the filtered dataset assigned above
            method = "api", # needs to be set in order to download from the api
            files.per.chunk = 10, # this should minimise prob. of corruption
            directory = "GDCdata") # save files to a seperate directory

TCGA_BLCA <- GDCprepare(query,
                        summarizedExperiment = TRUE) # Downloads also clinical data

TCGA_BLCA_DGEList <- DGEList(TCGA_BLCA)

keep <- rowSums(cpm(TCGA_BLCA_DGEList)>0.25) >= 217

TCGA_BLCA_DGEList <- TCGA_BLCA_DGEList[keep, , 
                                       keep.lib.sizes = FALSE]

TCGA_BLCA_DGEList <- calcNormFactors(TCGA_BLCA_DGEList, 
                                     method = 'upperquartile', 
                                     p = 0.75)

TCGA_BLCA_DGEList = estimateCommonDisp(TCGA_BLCA_DGEList, 
                                       verbose = TRUE)

TCGA_BLCA_DGEList$log2ps_counts <- cpm(TCGA_BLCA_DGEList$pseudo.counts,
                                             log = TRUE)

When I now subset my DGEList, log2cpm won´t be subsetted

tumor_tcga <- TCGA_BLCA_DGEList[, 
                                TCGA_BLCA_DGEList$samples$shortLetterCode == "TP"]
> dim(tumor_tcga$samples)
[1] 414 234
> dim(tumor_tcga$log2cpm)
[1] 18476   433
> dim(tumor_tcga$counts)
[1] 18476   414

I hope you can help me with that.

Thanks and best regards, Aljosch Leusmann  

I don't believe this is possible. The DGEList class was not designed to be a general-purpose container, and it will not recognize arbitrary user-specified fields. Perhaps there is an argument for adding direct support for SummarizedExperiment input to various edgeR functions - and indeed, this would clearly benefit a few of my packages (e.g., csaw, diffHic, cydar) - but this would involve a fair amount of work.

Thanks for the fast response.

Is it possible to transform a DGEList into a SummarizedExperiment?