Hi everyone,

I am using Illumina datasets from RNAseq with RobiNA and after the alignment, in the statistical analysis, when I try to run edgeR with tagwise dispersion the program just haltes.

It works fine if I choose auto dispersion but with tagwise the following error messages appears:

Loading required package: limma
warning messages:
1:package edgeR was built under R version 2.15.1
2:package limma was built under R version 2.15.1
Calculating library sizes from column totals
running estimateCommonDisp() on DGEList object before proceeding with estimateTagwiseDisp().
Error in t.default(object$counts): argument is not a matrix
Calls: estimateTagwiseDisp... splitIntoGroups -> lapply -> split -> t -> t.default
In addition: warning message :
In estimateCommonDisp(object):
There is no replication , setting dispersion to NA .
Execution halted

I am running it on windows
In the R directory I have the limma package

Does anyone knows what do i have to do?

best regards

pcastro

Dear Pedro,

as it seems, the data set you have entered does not contain replicates for each of
the different treatments you want to compare. If this is not true you have probably
organized the data in the "library configuration" step of the workflow in a way that
was misinterpreted as having no replication.

To clarify this: Each "box" you define represents one replicate (unlike in the microarray
analysis workflows). So, in case you have several .FASTQ files, each representing
one replicate RNA-Seq sample (or library), you would create a sample box for each of the
files individually, so that RobiNA "knows" that you have several replicates of the
respective condition or treatment.

If you really don't have replicates you cannot use tagwise dispersion estimation, because
this method requires replication to be able to assess the dispersion for each tag (gene)
individually and then subsequently uses bayesian statistics to "shrink" the values towards
the common dispersion. Without replication this does not work and hence the "auto"
setting will automatically choose to do only common dispersion estimation (without replication
the common dispersion is, as you can see in the error message, set to NA) if there
is no replication in the data.

I hope this answers your question - if you are unsure don't hesitate to aks again

cheers,
Marc