Normal skin is already full of "cancer" mutations

In 2015 a group sequenced 74 cancer genes to very high depth across 234 tiny biopsies of normal eyelid skin — about 5 cm² in total, from four adults aged 55 to 73 having blepharoplasty. The tissue was histologically and physiologically normal. The question was what a lifetime of sunlight had already written into it.

What they found

• Positively selected driver mutations — in NOTCH1, NOTCH2, NOTCH3, TP53, FAT1 and RBM10 — present in an estimated 18–32% of cells. The authors' own summary is "over a quarter."
• Roughly 140 driver mutations per square centimetre of skin.
• An overall burden of 2–6 mutations per megabase per cell, carrying a clear ultraviolet signature.
• On average 0.27 driver point mutations per cell — against 2.7 per cell in a cutaneous squamous cell carcinoma, roughly tenfold higher.

The epidermis kept working normally throughout. Whole fields of skin were carrying mutations that, found in a tumour, would be reported as drivers.

What this does and does not mean

It does not mean these mutations are harmless, and it does not mean the tissue is pre-cancerous in any clinically actionable sense. What it establishes is narrower and more useful: finding a driver mutation in a sample is not by itself evidence of cancer. Context — how many, in what combination, in what tissue architecture — is doing most of the work.

Why a decision-support project cares

Because "a driver mutation was found" is a sentence that arrives in clinical reports and gets read as more than it is. A knowledge base that maps biomarkers to treatment has to encode which findings are actionable in which disease context, rather than treating any hit in a cancer gene as meaningful on its own.