Hi, I am quite new to R and I have been trying to use a dataset from the GEO database (GEO104438) that is interesting to my PhD work. The file that I have been able to download and use in R, is a table with ID references and then the normalised signal intensity values for all of the donors (and cell types). I have put a screenshot of the table for reference.

I am hoping to try look at whether there are any genes that are differentially expressed depending on what cell type they are (as they are both derived from a common precursor). So for each donor I have 2 different cell types. Due to this, I was initially trying to tell R that my data is paired but I am stuck.

I have been reading online and trying to use the limma guide but I am completely confused as all of the examples I have seen online are not in a similar format to this dataset.

I would appreciate any help!

I would use GEOquery to get the data.

> dat <- getGEO("GSE104438")[[1]]

## now you have to get rid of some extra cruft
> names(fData(dat))
 [1] "ID"                   "SPOT_ID"              "CONTROL_TYPE"        
 [4] "REFSEQ"               "GB_ACC"               "GENE"                
 [7] "GENE_SYMBOL"          "GENE_NAME"            "UNIGENE_ID"          
[10] "ENSEMBL_ID"           "TIGR_ID"              "ACCESSION_STRING"    
[13] "CHROMOSOMAL_LOCATION" "CYTOBAND"             "DESCRIPTION"         
[16] "GO_ID"                "SEQUENCE"    

## that's too much
> fData(dat) <- fData(dat)[,c(1,6:8)]
## now parse the phenotypic data. It's usually a mess
> pData(dat)$title
 [1] "ATDC Donor 1"    "ATDC Donor 2"    "ATDC Donor 3"    "ATDC Donor 4"   
 [5] "ATDC Donor 5"    "ATDC Donor 6"    "MoDC Donor 1"    "MoDC Donor 2"   
 [9] "MoDC Donor 3"    "MoDC Donor 4"    "MoDC Donor 5"    "MoDC Donor 6"   
[13] "MoMacro Donor 1" "MoMacro Donor 2" "MoMacro Donor 3" "MoMacro Donor 4"
[17] "MoMacro Donor 5" "MoMacro Donor 6"

## that will work
> samps <- as.data.frame(do.call(rbind, lapply(strsplit(pData(dat)$title, " "), "[", c(1,3))))
> names(samps) <- c("Cell","Subject")
> samps$Subject <- factor(samps$Subject)
> samps
      Cell Subject
1     ATDC       1
2     ATDC       2
3     ATDC       3
4     ATDC       4
5     ATDC       5
6     ATDC       6
7     MoDC       1
8     MoDC       2
9     MoDC       3
10    MoDC       4
11    MoDC       5
12    MoDC       6
13 MoMacro       1
14 MoMacro       2
15 MoMacro       3
16 MoMacro       4
17 MoMacro       5
18 MoMacro       6

## Much better

## now it's straightforward
> design <- model.matrix(~0+Cell + Subject, samps)
> contrast <- makeContrasts(ATDC - MoDC, ATDC - MoMacro, MoDC - MoMacro, levels = design)
> fit <- eBayes(contrasts.fit(lmFit(dat, design), contrast))

That's all pretty condensed code. I leave it to you to read the limma User's Guide to understand why I did what I did, and to play around with the code and data to understand what each line of code is doing.