I need help figuring out the best way to do the following experiment.  I did some previous work which separated the samples into 4 groups based on data other than RNASeq.  I then used EBSeq to look for differentially expressed genes in those groups, and I found 6 genes.  Now I have a fresh set of samples and I want to see if I can go backwards where I put them into four classes based on the previously discovered differential expression profiles and see if those classes have the same properties that I had originally used to make my four groups.  Sort of a validation experiment.

I have a good bit of experience with machine learning, and I know generally how to do something like this, but I also know RNASeq data has some properties that complicate things, such as being very non-gaussian.  Does anyone know of a bioconductor package designed for this task?

Thank you!