I have raw rna seq data in the following form:
Genes P1_stage1 P1_stage2 P1_stage3 P1_stage4 P1_stage5 ............ P7_stage7
Where P1_stage1 denotes person1 at stage1 of the Virus. And the dataset is populated with gene expression counts. We have on average 7 stages per individual, where a 'single' infection of the Virus lasts 3-4 stages.
I have previously used camera() to obtain values for the significance of particular genesets however, I would like to obtain the significance of these genesets with respect to the individual person and whether it is the persons 1st,2nd, (or 3rd for one patient) infection.
Problem is, previously when using camera(), I have used voom, eBayes and lmFit objects and now, that I have in some instances only 1 or 2 samples that I would wish to do the gene set testing on, I am getting many errors along the line as a result.
EDITED HERE::: The type of errors I get are generally as such:
Error in approxfun(l, rule = 2) : need at least two non-NA values to interpolate.
I have found a post that suggests trying, what they called, 'blind' design, where the matrix is modeled as such: model.matrtix(~1, data = data). When I use this it works, but the method is not convincing. Does anybody have any experience doing it this way? What are the implications of such and I'm not really convinced but, is there any other way around it?
If anybody can offer some guidance on how to proceed, that would be very helpful.
Thank you all for your time,