Hypermutation Gives Weak Cancer Drivers a Head Start, Study Finds

Some of the most puzzling patterns in cancer genetics are now being explained not by what tumours are selected to do, but by how fast and how sloppily they mutate. A new computational study published in PLOS Computational Biology by Michael D. Nicholson and Ian Tomlinson shows that hyper-mutational processes alone can account for why certain weak cancer driver mutations, particularly atypical variants of the KRAS gene, come to dominate some tumour types. The finding challenges a long-standing assumption that the driver mutations observed in a tumour must reflect the strongest possible oncogenic advantage. Instead, the researchers demonstrate mathematically that simply cranking up the mutation rate per cell division can allow sub-optimal drivers to win the evolutionary race, without any change in selection at all.

KRAS is one of the most frequently mutated oncogenes in human cancer, acting as a molecular switch that locks cells into a growth-promoting state. In most colorectal cancers, the same handful of canonical KRAS variants appear again and again, particularly substitutions at codons 12 and 13 that interfere with the protein’s ability to switch itself off. These canonical mutations are potent drivers, and their recurrence has traditionally been read as evidence of strong positive selection. Yet in hypermutant cancers, tumours burdened with extraordinarily high mutation loads, a strikingly different picture emerges. Atypical KRAS variants, mutations at unusual positions or of unusual types, appear far more often than their oncogenic strength would seem to justify.

This paradox has troubled cancer evolutionary biologists for years. If a weak driver confers less of a growth advantage than a canonical one, and if both are available to a tumour, conventional evolutionary logic predicts that the stronger driver should prevail and outcompete the weaker alternative. The persistence of weak drivers in hypermutant tumours therefore demanded an explanation. One intuitive hypothesis was that these tumours might somehow favour weak drivers through differential selection, perhaps because the cellular context of a hypermutant cancer changes what mutations are useful. Nicholson and Tomlinson set out to test whether such an appeal to altered selection was actually necessary.

Their approach was to build mathematical models of tumour evolution that track the appearance and expansion of driver mutations over the course of a cancer’s development. The models incorporate the basic currency of cancer genetics: cell divisions, mutation rates, the probability that a given division produces a driver mutation, and the selective growth advantage each driver confers. By varying these parameters, the researchers could ask under what conditions weak drivers would be expected to dominate the final tumour population. The answer they found was deceptively simple. When the mutation rate per cell division is elevated, further driver events occur so rapidly that the dynamics of tumour evolution are fundamentally altered, and weak drivers can rise to dominance even in direct competition with stronger alternatives.

The intuition behind this result is worth unpacking. In a tumour with a normal mutation rate, driver events are rare. A cell that acquires a strong driver mutation has time to expand into a large clone before the next driver arises, so the strongest available driver tends to sweep through the population. But in a hypermutant tumour, driver events arrive in quick succession. A cell carrying a weak driver may acquire additional driver mutations before any stronger competitor has had the chance to establish itself. The weak driver effectively gets a head start, and the sheer speed of subsequent mutation means that the tumour’s evolutionary trajectory is shaped more by the order and timing of events than by the fine ranking of selective advantages. Selection still matters, but the race is run so fast that the starting position can matter more than the finishing speed.

The researchers also found that this effect can be amplified by mutational biases, systematic tendencies of a mutational process to produce certain nucleotide changes in certain sequence contexts rather than others. Mutational processes are not random. Ultraviolet light, tobacco smoke, defective DNA repair enzymes and defective proofreading polymerases each leave characteristic fingerprints in the genome, preferentially mutating particular bases in particular neighbourhoods. If a hypermutational process happens to target the sequence contexts required for atypical KRAS variants, while rarely generating the contexts needed for canonical ones, then the supply of candidate weak drivers is inflated relative to the supply of strong ones. Combined with an elevated mutation rate, this bias can be sufficient to make atypical drivers the dominant species in the tumour.

To test whether this theoretical framework matched reality, the team focused on a specific and clinically important example: colorectal cancers with mutations in POLE, the gene encoding the proofreading domain of DNA polymerase epsilon. When this proofreading function fails, the enzyme makes copying errors at a greatly elevated rate, producing some of the highest mutation burdens seen in any human cancer. Nicholson and Tomlinson quantified the mutation bias of the POLE-mutant process for a range of KRAS driver variants, comparing the sequence contexts generated under POLE deficiency with those generated under non-hypermutant mutational processes. They found that the combination of the POLE-specific bias and the elevated mutation rate was sufficient to explain the observed enrichment of atypical KRAS drivers in these tumours. No special selective regime needed to be invoked.

The model’s predictive power extended beyond POLE-mutant cancers. Mismatch repair deficient colorectal cancers, another hypermutant class arising from a different DNA repair defect, also show an elevated prevalence of atypical KRAS drivers. When the researchers applied their model to this setting, its predictions were consistent with the observed variant frequencies. This is a notable result because POLE deficiency and mismatch repair deficiency generate mutations through different chemical mechanisms with different sequence biases, yet both produce the same qualitative outcome: weak, atypical drivers rising to prominence. The convergence suggests that the underlying principle, hypermutation reshaping the balance between mutation supply and selection, is general rather than specific to one repair pathway.

The broader implication of the study is a clarification of how mutation and selection interact during tumorigenesis. Textbook accounts of cancer evolution often treat driver mutations as the winners of a pure selection contest, with the fittest mutations sweeping to fixation. This work shows that in hypermutant tumours, the mutational process itself can act as a filter that determines which drivers are even available in meaningful numbers, and that the tempo of mutation can override fine-grained differences in selective advantage. Differential selection across cancer types need not be invoked to explain why different tumour classes carry different driver repertoires. The variation may simply be a consequence of how fast each tumour mutates and what kinds of mutations its particular defect produces.

There are also practical consequences for precision oncology. Atypical KRAS variants have historically been poorly characterised compared with their canonical counterparts, and their functional consequences are often uncertain. If many of these variants arise as passengers or weak drivers propelled by hypermutation rather than as potent oncogenic signals, this has implications for how they are interpreted in clinical sequencing reports and for which variants become targets of drug development. More broadly, the study adds to a growing appreciation that mutation rate is not merely a background parameter of cancer evolution but an active force that shapes which evolutionary paths tumours take. Understanding the interplay between mutational processes and selection, the authors conclude, is essential for explaining the diversity of driver mutations observed across human cancers, and mathematical modelling of the kind employed here offers a way to disentangle these forces that raw genomic data alone cannot provide.

Subject of Research: Evolutionary dynamics of weak KRAS driver mutations in hypermutant colorectal cancers

Article Title: Hyper-mutational processes provide a head-start for weak cancer drivers: Explaining atypical KRAS variants

Article References: Nicholson, M. D., & Tomlinson, I. (2026). Hyper-mutational processes provide a head-start for weak cancer drivers: Explaining atypical KRAS variants. PLOS Computational Biology, 22(10), e1014858. https://doi.org/10.1371/journal.pcbi.1014858

Image Credits: AI Generated

DOI: 10.1371/journal.pcbi.1014858

Keywords: KRAS, hypermutation, POLE, colorectal cancer, mathematical modelling, cancer evolution, driver mutations, mismatch repair deficiency, mutational bias, PLOS Computational Biology, tumorigenesis, selection