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.