Hello,
I'm trying to perform a DE analysis on NanoString data. Using NanoStringDiff 1.16 I can get the example data that comes with the package to run in ~10 minutes following the package vignette:
> directory <- system.file("extdata", package="NanoStringDiff", mustWork=TRUE)
> path<-paste(directory,"Mori.csv",sep="/")
> designs=data.frame(group=c("Normal","Normal","Tumor","Tumor"))
> NanoStringData=createNanoStringSetFromCsv(path,header=TRUE,designs)
> pheno=pData(NanoStringData)
> group=pheno$group
> design.full=model.matrix(~0+group)
> contrast=c(-1,1)
> NanoStringData=estNormalizationFactors(NanoStringData)
> result=glm.LRT(NanoStringData1,design.full,contrast=contrast)
However, when I load my own data, all steps run (and based on the output appear to run correctly) until I get to the final command:
> designs=data.frame(group=c("Control","Control","Group1","Group1"))
> NanoStringData=createNanoStringSetFromCsv(mydata.csv,header=TRUE,designs)
> pheno=pData(NanoStringData)
> group=pheno$group
> design.full=model.matrix(~0+group)
> contrast=c(-1,1)
> NanoStringData=estNormalizationFactors(NanoStringData)
> result=glm.LRT(NanoStringData,design.full,contrast).
It hangs without producing any output. I gave it 20 hours before I had to kill it. Any ideas what could be making it take so long?
I have re-checked the formatting of my input data and it matches exactly what is detailed in the vignette (a csv count table with the three first columns being "Code.Class", "Name", and "Accession". I have also tried this with versions 1.16 and 1.14 of NanoStringDiff and I have tried it on both Windows and Linux platforms.
My dataset has 768 features and 15 samples so I didn't think it should take much longer than the example.
Is there perhaps a verbose option to get some progress output from the glm.LRT command?
You will have to provide more than that. For anybody but you, what you have told us is that you did some stuff and it didn't work, which isn't much to go on.
You might read the posting guide, and edit your post to provide enough information that someone might be able to help you.
I have added what I can but since the command I mentioned before isn't producing an error or any progress output of any kind I also don't have much to go on.
Well, from what you have shown us, it appears that you are using the same design and contrast matrix for your data as you used for the example data, which seems suboptimal.
No. I load my own data and run all the same steps on my data. The same commands, still two groups, but on my own data to generate all the matrices for the actual dataset to be analyzed.
You have two groups, made up of two samples each? I thought you said you have 15 samples?
I've tried it with 15 samples, a single pairwise compaison (2 samples in control, 3 samples in group1), and a second pairwise comparison with 2 samples in control and 2 samples in group1. This example code is from the last attempt. All hang for >3 hours.
You'll probably need to provide the maintainers some self-contained code that hangs so they can debug.