In a way, sure. A major roll of the immune system (especially T cells) is to monitor for unusual cells. Some pre-cancerous cells display unusual markers that are recognized and killed by T cells. Those that evade T cell recognition go on to be a problem.

The major problem is still how do you target cancer cells? If you can find something that makes them unique, say a cell surface receptor, or an unusual metabolic process, there are lots of ways to target them for removal. It's just hard to uniquely identify cancer cells. Right now, most chemotherapy drugs target the phenotype of "fast replication" which really isn't too selective.

> It's just hard to uniquely identify cancer cells.

How about performing whole-genome sequencing on a tumor, and then targeting the differences with the patient's natural genome?

That is a potential approach. However, the things that cause cancer are typically dysregulation of self proteins. These are usually things that stop and start cell growth and replication. In these cases, it is not totally clear in these cases what can cause recognition (perhaps overabundance causes higher presentation on the outside of cells and maybe enhanced recognition by some T cells can cause killing--I am speculating as I'm a couple years out of date on the literature).

Other cancers there may be a mutation in one of these genes that makes it more or less effective and lead to cell cycle deregulation. Those are good targets for T cell killing.

In any case, it is probably a more effective approach in cases where we understand potential epitope targets to directly expand T cells that can do the killing rather than having some intermediary in the mix.

Tumor cells encode the same genes as non-tumor cells.

The difference is more likely to be in the expression level or epigenetic (de)repression of a particular gene than the presence or absence of that gene. Organisms don't tend to keep around genes that only serve to form tumors.

A slightly better approach would be to use RNA-sequencing to look for mRNA signatures in tumor vs normal cells.

Wouldn't work reliably due to cases of genetic chimerism, though that could possibly be tested for. Though, in several cases I've read about, there was no consistent place to get samples to test for mismatched DNA.

That sounds way more complicated (expensive) than using a minimalist signature to identify cancerous cells.

With desktop gene sequencers becoming a near commodity product I’m not so sure.

There are a companies that currently focus on this area specifically by leveraging the ever dropping costs and capabilities of gene sequencing by developing targeted immunotherapy solutions that can target the specific genetic makeup of a tumor from a biopsy.

This is all based on the assumption that our understanding of in vivo processes is as good as our understanding of in silico processes. It isn't. Not even close.

But that's also true for mainstream/existing therapies.

My thoughts exactly! Here's how it might work :

Get a construct into the cell that is only transcribed if specific pathways are overly active, effectively testing that cells "cancer status". This is the hard part.....

Then have it translate something pathogenic-looking that gets presented on the surface.

So now the question is, how do you design some kind of promoter sequence that looks for, say, tp53 mutations or something similar?

I like your silver-lining take on this. It wouldn't be bad low-hanging fruit to look at if someone hasn't already done that study.