What Is The Proper Control Sample For A Tumor Rna-Seq Experiment?

Question

[Hmm]

If i am doing RNAseq analysis of a tumor sample. What do i use to compare the tumor(suppose pancreas) values to? Would any tumor pancreas control from any patient work or i need to have the same patient normal RNAseq as well? Sorry if this sounds ignorant but this is the first time i am trying to analyze RNAseq.

[Aaron Statham]

What hypothesis are you testing?

[Hmm]

@Aaron Statham For a start i was just thinking of doing differential expression.The other things possible are alt splicing,novel exons,mutations...!

Answer 1

[Sean Davis]

To follow up on Aaron's comment, good science does not end when one begins to discuss technologies such as RNA-seq. Whether or not one uses RNA-seq, microarrays, PCR, or a crystal ball, it is important to think of the biologic questions that you want to answer. Layer on that a significant helping of reality (try getting matched normal associated with a brain tumor, for example), and you can begin to answer your question, which needs to be experiment-specific. Keep in mind that you may have to ask a significant number of questions of your bench/clinical collaborator and also read the literature and that, sometimes, there is NOT a well-accepted answer as to what the appropriate normal is (some sarcomas, for example, for which the cell-of-origin is not known).

Answer 2

[Zhidkov]

I think the best control is to use normal corresponding tissue of same individual (if possible). In that case you can check if variation in gene expression or splice variance that you see in tumor is actually tumor specific and not tissue or case specific. If you can't use corresponding normal tissue, again good idea is to use other normal/control tissue from same individual (but you have to keep in mind that some variation in expression and splicing can be tissue specific). Tissue specific expression you can check on other samples or databases (such as http://www.ebi.ac.uk/asd/).

Ilia

[Sean Davis]

If I understand the comment above, I have to disagree a bit here. The variation between tissues in an individual is HUGE compared to the variation between individuals within the same tissue. I agree that, in general, the best bet is to get normal tissue from the case individual herself. Lacking that control, though, the next best bet is to get controls from the normal tissue from which the tumor is thought to arise. Note that this rule-of-thumb applies to gene expression (rna) but not to SNVs or other germline or somatic variation.