Relevance and validation are separate issues. qPCR validation is to add or subtract support that the results you see in your arrays (or RNA-Seq) are due to the biology and aren't technical artifacts. Relevance, as you said, is about if a given DE result plays a role in the phenotypes of what you compared. You can have a qPCR validated gene with strong DE that doesn't have a role in your biology of interest.
You still want to do functional analysis on the data you have, it will greatly aid in interpreting your data and navigating follow up. You could also compare functions enriched/depleted in colon cancers versus other cancer in other tissues. Make sure you also compare normal colon tissue against normal tissues from other cancers. You'll want to be careful about getting distracted by false positives caused by tissue-specific patterns of gene expression.
As a follow up, there might not be anymore tissue, but is there any left material from RNA extractions or cDNA? You don't want to or need to extract fresh RNA for qPCR validation. I guess if you have zero tissue left, you could look at previously published gene expression studies and see how well your results mirror them. Another option would be to collect more tumor samples and do your qPCR there. You really need the validation, some journals simply require it flat out, and I can't imagine many editors would let you through without it.
This probably isn't the answer you want: To really know if your gene is important to colon cancer, you have to do the wet work.
Make a list of potential targets from genes upregulated in cancer. Then knock them down/out in a colon cancer line and check for differences in phenotype. You should do the same knockout/down in a normal colon cell line (I think ATCC has one) and an unrelated cancer as controls (along with mock knockdown/outs). If a KO causes significantly different changes in phenotype in the colon cancer line, you might be on to something.
On the other side of the coin, you should consider looking at genes that are strongly down regulated or not expressed in your tumor samples. Instead of knockouts, you would want to express these in your tumor cell line.
You could also look for drugs that inhibit things that are upregulated in tumors.
The only draw back here is that the normal/tumor cell lines might not behave like real tissue. You should first use qPCR to see if your targets are still DE before getting knee deep into follow up experiments.
Dear Joe,
thanks for the comprehensive answer, really many things to consider !! Based on the first part of your analysis: because some of our collaborators performed the microarray experiments before some years, and i recently analyzed them, i doubt if any RNA material left from the RNA extraction, and if is possible at any way to find new material for these patients, because the surgeries as i said where performed before years, but i will search it anyway.
Now, regarding your second part of your answer: one similar notion of your suggestions, is to use some drug repositioning methodologies, and find any colorectal cancer cell lines, and retreive the gene expression profiles for my up-regulated genes--i.e. my upregulated genes found downregulated after various drug pertubations.
Also, one naive thought is for my top results, for istance the top10 upregulated genes by log-fold change, or adjusted p-value, to quire them in some other very interesting tools, like the cancer genome atlas browser and BioXpress. Although both deal with RNA-seq data, would be an additional point any corcondance found in CRC cancer for these genes.
I've been in this situation before, if there is RNA still, hopefully it was stored in RNALater or something. I've not had luck with very old RNA in the past.
Not sure what drug repositioning methodologies are. Seeing that your target is downregulated after treatment with a drug just means that gene was downregulated after exposure to the drug. If you see a different phenotype in normal cells, that might be an interesting result. However, that isn't as strong of a link between over expression of that gene and the observed phenotype. This might be interesting in a network approach, but you'd need whole transcriptome estimates of transcript levels for each drug treatment to really figure it out.
Direct inhibitors of upregulated targets or siRNA in normal and tumor cell lines, with the appropriate assessment of phenotype, would be the most direct approach to providing evidence to your claim that your target may play a role in colon cancer.
Thank you again for your feedback---for drug repositioning, i meant something like this very interesting tool--http://amp.pharm.mssm.edu/L1000CDS2/#/index
Very briefly, as my patients from which my microarray datasets resulted had taken no therapies--if i found any resulted colon cancer cell lines, that a subset of my input up-reg DE genes found down-reg in these, i could have a putative hypothesis. However, i catch your notion about the necessity of capturing the whole transcriptome estimates--then of course, maybe the same drug from the drug repositioning tested on the same up-genes in a normal colon cell line, could further give a stronger significance in any interesting results...