What to do: RNAseq with multiple samples of the same disease

[cnmartinez11]

Here is the scenario: I have RNAseq data from seven tumor samples. I've put them through the cufflinks pipeline including both cuffdiff and cuffnorm. For cuffdiff, I treated each sample as its own condition with no replicates. I'm aware this cannot produce statistically significant results, but beyond that differential expression isn't really what I want anyway as all these samples derive from the same disease.

Broad question: Is there anything meaningful that can be done in the given scenario?

More specific:

• If I have a list of a few genes that I'm interested in. As I understand, I cannot simply compare (within each sample separately) the FPKM of a gene of interest with other random genes (or e.g. an average across all genes). Can I do anything with this list?
  
  
• I have access to other RNAseq data from public databases. Does it make sense (for exploratory purposes anyway) to do differential expression between my tumors and e.g. breast cancer tumors? Given tissue- and individual-dependency of expression, I'm guessing not, but I'm wondering if any of you had more insight.

I'm grasping at straws here because of the lack of planning in this experiment (there is normal and tumor DNA but not matched RNA!). Thanks for your help.

[dpryan]

Is there anything meaningful that can be done in the given scenario?

Maybe cluster the samples to see if there are any obvious subtypes of samples?

Can I do anything with this list?

RNAseq data would have to undergo a lot of normalization if you wanted to directly compare genes within a sample. I suppose you could get the rankings of the genes in the list and do something with that. Generally a hypothesis at the beginning is helpful.

Does it make sense (for exploratory purposes anyway) to do differential expression between my tumors and e.g. breast cancer tumors?

I wouldn't bother unless, the results would be too confounded to allow drawing any real conclusions. Further, I wouldn't personally use the results to suggest that someone in the lab spend time/money on follow-up experiments.

[cnmartinez11]

Great! Thanks for your guidance.