Copy number alteration (CNA)

It separates two explanations that are usually run together. Some of the difference in prostate cancer outcomes by race is in the tumour genome and persists when access to the same centre is held constant, and some of it tracks with income rather than with ancestry, so equalising access alone would not eliminate the gap.

Lung cancer in never-smokers is not smokers' lung cancer with the smoking removed; it is a different set of diseases with a different clock. The slow-growing piano subtype in particular is the argument that a screening test aimed at never-smokers would need to look for something other than what low-dose computed tomography was built to find.

It shows that the driver frequencies quoted in Western guidelines are population statistics rather than facts about the disease, which matters for how many patients anywhere are expected to benefit from a given medicine.

It explains why single-sample classifiers disagree on about one tumour in eight and why KRAS allelic imbalance and GATA6 copy number are being read alongside expression.

It shows that the commonest driver event in advanced prostate cancer is invisible to the panels used to test for it, and that the shape of the structural damage in a genome tells you which repair pathway failed, which is information a mutation list does not carry.

It is the reason a prostate cancer fusion frequency quoted without an ancestry is unsafe. In this cohort the founder event that defines almost half of Western tumours is uncommon, and the fusion-negative, CHD1-deleted route dominates instead.

It validates a clinical definition against a molecular one, which is unusual and useful: a man whose disease behaves like small cell carcinoma can be treated as such even when his biopsy does not look like it, because the underlying genotype is the same.

Why small-cell lung cancer has no targeted therapy in the conventional sense: it is built from the loss of TP53 and RB1, and the drugs that transformed non-small-cell disease inhibit gains rather than restore losses. The NOTCH result pointed at DLL3 and, eventually, at tarlatamab.