I have to analyze an RNA-seq dataset. Goal is to compare two groups - say case and control. The issue is that there is only one sample per group. In a normal situation , I would not proceed with the analysis as n=1 is not really a "group", its not statistically justifiable, results cannot be generalized.

But this data is on cell line from a real patient with a disease. I will circle back with the investigator to see if its possible to generate more data. But in the event that its not possible to get more data, would voom-limma be a good tool to try (with all the caveats mentioned above). Thanks.

limma-voom cannot analyse data without replicates. The only option is to use edgeR with a preset dispersion parameter. See "What to do if you have no replicates" in the edgeR User's Guide:

• https://bioconductor.org/packages/release/bioc/vignettes/edgeR/inst/doc/edgeRUsersGuide.pdf

It is very easy. If counts is your matrix of read counts with two columns corresponding to case and control, then you can do a DE analysis by:

library(edgeR)
y <- DGEList(counts, group=c("control","case"))
y <- normLibSizes(y)
et <- exactTest(y, dispersion=0.2)
topTags(et)

Of course, the dispersion setting here is arbitrary and having replicates would be infinitely better. Nevertheless, the above analysis with any positive value for the dispersion is vastly better than assuming Poisson variation, as very many papers in the literature have done in similar situations.

The value of 0.2 that I have chosen is fairly conservative. Good quality RNA-seq data on a cell line should be less variable than that.