In science – especially biology – what you find is no more important than where you look.
The oncogene paradigm of cancer genesis — that mutations to key genes cause cells to become cancerous – has reigned supreme for several decades now. It is a driving factor in the whole notion of personalized medicine, which in turn now dominates cancer drug development efforts: tumors are sequenced, driver mutations are identified and appropriate treatments are prescribed.
This paradigm has an enormous amount of data supporting it. Normal cells can be transformed into cancer-like cells by oncogenes. Conversely, cancer cells typically bear mutant or amplified oncogenes.
There are problems with this story. One is that only a small minority of cancers exhibit genetic signatures that can be used to choose therapies. The other is that even this small minority often fails to respond more strongly to targeted therapies than to conventional therapies. And our attempts to predict cancer risk from genomic signatures are almost laughably lame in most cases. Outside of a few high-penetrance but low-frequency mutations such as BRCA, genomic algorithms predict absolute risk no more than a few percent better than population-wide incidence. Although statistically significant, knowing that your lifetime risk of colon cancer is 2% rather than 1% is not useful information. The bacon industry has nothing to fear from genomics.
Our ability to generate information generally outruns our ability to interpret it. And our medical system is incentivized to find “abnormalities” which are eligible for treatment and thus compensation. We know that MRI scans of healthy individuals with no back pain are just about as likely as to show abnormalities as those reporting significant pain. In other words, an abnormal MRI may have nothing at all to do with your back disease.
And it is beginning to look like the same may be true of cancer genes. As sequencing technology has gotten cheaper and more sensitive, requiring fewer cells, researchers have begun sequencing healthy tissues and looking for mutations associated with cancer in smaller and smaller samples. It turns out that these mutations are not rare at all. In fact, they are more predictive of aging than they are of cancer.
From Aging and the rise of somatic cancer-associated mutations in normal tissues
Not only are cells prone to accumulate mutations with age and additional divisions, but cells that acquire growth-promoting mutations may well come to dominate a tissue — and still not be cancerous. Worse, some genes thought to be cancer drivers are more prevalent in healthy tissues than in cancerous ones.
Much of our fascination with DNA and genomics stems from our desire for order and just-so stories. We have ascribed god-like properties to DNA, thinking it a master planner which directs our fates. We study it as if a holy text, looking for powers that we can use to achieve outcomes to our liking.
Maybe DNA is just one biochemical among many. It is rich in informational content, but the meaning of that information is highly dependent on context. Genes that promote cancer in one individual, or one tissue, or one life history or environment may have no cancerous tendencies at all in other circumstances.
Genetic determinism has long been debunked as a rationale for outcomes at a population level (that is, as a justification for racism). It appears to have little value at the level of individual fates. And now it seems we can’t even apply it with confidence to tissues and cells. Cancer biology may need a new story.