no significant genes with differential expression due to 1 vs 1 sample comparison

Question

[jolin0701-dy]

I have 1 sample for control, 1 sample for treatment.

After the pipeline, tophat>cufflinks>cuff merge> cuffdiff

no significant genes with differential expression.

But there are still many genes with log2 fold change >1 or <-1.

I know the sample is too few for normal cuffdiff but since there are human samples and this is what I have.

So any suggestions for the analysis?

Thanks~

[GenoMax]

If that is all you have (and no possibility of getting replicates/more data) then you will have to make do with the data in hand.

Yes there is zero "statistical" support for the differences you found but someone (or you) could design/do additional experiments (e.g. qPCR) to see if a few of the top ranked genes you are seeing are real/reproducible.

[WouterDeCoster]

Yes indeed, perhaps I was a little too brief in my reply.

An additional option would be (depending on the comparison you want to make, e.g. diabetic liver vs healthy liver) to use publicly available data from micro array or RNA-seq with the same phenotype and tissue, e.g. from GEO (http://www.ncbi.nlm.nih.gov/geo/) or Genevestigator (https://genevestigator.com/ - partially free, partially commercial).

But this of course depends whether your phenotype of interest has already been studied in a comparable tissue.

Answer 1

[WouterDeCoster]

When you don't have replicates there is no way to identify significantly deferentially expressed genes, a one vs one comparison is completely invalid by any method.