Driver and passenger mutations: the refined somatic mutation theory

The claim. A typical adult solid tumour carries tens of thousands of somatic mutations, but only two to eight of them, in a few hundred recurrently mutated genes, confer a growth advantage. Drivers fall into a dozen signalling and housekeeping pathways; Kinzler and Vogelstein separated gatekeepers (which directly control growth) from caretakers (whose loss speeds mutation). Passengers are neutral hitchhikers, useful as a record of the tumour's exposure history and as neoantigens, but not as targets.

Who and when. Kinzler and Vogelstein set out gatekeepers and caretakers in 1997; Stratton, Campbell and Futreal's 2009 review The cancer genome defined drivers against passengers as sequencing scaled; Vogelstein and colleagues' 2013 Cancer genome landscapes described the 'mountains and hills' of driver frequency and estimated the number of drivers per tumour; Alexandrov, Nik-Zainal and Stratton read mutational signatures from the passengers in 2013; Martincorena and colleagues showed in 2017 that the ratio of protein-changing to silent mutations identifies drivers under positive selection and that most passengers are effectively neutral.

Evidence for. The same genes recur across thousands of tumours far more often than chance allows, and the frequency spectrum has a small number of very common drivers and a long tail of rare ones. Drugs against drivers work regardless of the passengers. Tumour-agnostic approvals by driver (larotrectinib for NTRK fusions, pembrolizumab for mismatch repair deficiency) followed directly from the model.

Evidence against and limits. Passengers are not entirely neutral: McFarland and colleagues argued in 2013 that mildly deleterious passengers accumulate and slow tumour growth, and passengers supply the neoantigens that make a tumour visible to the immune system. Driver lists depend on the statistical method and on tissue context; a BRAF V600E mutation is a curable driver in melanoma and a poor-prognosis marker that resists BRAF inhibitors alone in colorectal cancer. The normal-tissue sequencing studies found canonical drivers under positive selection in healthy skin and oesophagus, so 'driver' describes a fitness advantage for the clone, not a guarantee of cancer. Many drivers (TP53 loss, MYC) are still not druggable.

Predictions that held or failed. Held: basket trials of driver-matched drugs produce responses across histologies; profiling finds an actionable driver in a substantial minority of advanced tumours; resistance mutations appear at predicted positions in the drug target. Failed: presence of a driver does not guarantee a response; whole-genome sequencing of every patient has not translated into a benefit for most; the 'long tail' of rare drivers remains largely untargeted.

Therapies that came from it. Comprehensive genomic profiling as routine care in lung, colorectal and other cancers, basket and umbrella trials, tumour-agnostic approvals, and the use of mutational signatures (homologous recombination deficiency, mismatch repair deficiency) to choose PARP inhibitors and immunotherapy. The model feeds clonal evolution, which explains how drivers are ordered in time and how resistance emerges.

Status: established. The driver and passenger distinction is settled in principle; the open questions are about context (which drivers matter in which tissue and at which age) and about how many rare drivers remain to be found.