Every tumor carries thousands of DNA mutations, but only a handful actually push cells toward uncontrolled growth. Those few are called driver mutations. The rest, often making up the vast majority of a tumor’s mutational catalog, are passenger mutations: changes that were present in the cell when it divided or arose during replication but did not themselves contribute to cancer’s progression. The distinction sounds clean, but in practice the line between the two is far blurrier than textbooks suggest, and that ambiguity has real consequences for how cancers are treated and how new therapies are developed.
What Makes a Mutation a Driver
A driver mutation gives a cell a growth advantage. It might activate a gene that tells the cell to divide faster, or it might disable a gene whose job is to stop damaged cells from multiplying. These two broad categories, oncogenes and tumor-suppressor genes, are the classic targets. Oncogenes get switched on or amplified through mutations, copy-number gains, or chromosomal rearrangements, while tumor-suppressor genes lose their protective function through inactivating mutations or deletions.
1PubMed Central. Exploring the Genetic Orchestra of Cancer: The Interplay Between Oncogenes and Tumor-Suppressor GenesIn evolutionary terms, a driver confers a fitness advantage to the cell that carries it. That cell outcompetes its neighbors, divides more often, and its descendants come to dominate the local tissue. A passenger, by contrast, has no meaningful effect on fitness. It just happens to be along for the ride in a cell that was already expanding for other reasons.
2PubMed Central. Tumor evolution: Linear, branching, neutral or punctuated?Mathematically, the selective advantage a typical driver provides is surprisingly modest. Modeling studies estimate that each driver mutation increases a cell’s net growth rate by only about 0.4%, a tiny edge that nonetheless compounds over years and thousands of cell divisions to produce a clinically detectable tumor.
3PubMed Central. Accumulation of driver and passenger mutations during tumor progressionHow Passengers Pile Up
Every time a cell copies its DNA and divides, it introduces a small number of errors. Most of these errors land in stretches of the genome that have no immediate effect on cell behavior. Some change a protein slightly but not enough to matter. When a driver-bearing clone expands, it drags along every passenger mutation that was already in that cell, plus new ones acquired during each subsequent round of division. The result is a snowball effect: a tumor with a few drivers can carry hundreds or thousands of passengers.
The pattern of passenger mutations is not random noise, though. It reflects the specific mutational processes that were active in the tissue. Ultraviolet radiation leaves a distinctive signature in skin cancers. Tobacco smoke leaves a different one in the lung. Endogenous processes like those involving APOBEC enzymes, which normally help fight viral infections but sometimes mistakenly edit the cell’s own DNA, leave yet another pattern.
4PubMed Central. Mutation signatures of carcinogen exposure: genome-wide detection and new opportunities for cancer prevention These mutational signatures are now routinely used to reconstruct which exposures shaped a given tumor’s genome, and they help researchers separate the background hum of passengers from the signal of drivers.
How Researchers Tell Them Apart
Distinguishing drivers from passengers is one of the hardest problems in cancer genomics. You cannot look at a single mutation in isolation and know which category it belongs to. Instead, researchers rely on statistical and computational strategies that look for patterns across many tumors.
One widely used approach compares how often a gene is mutated in tumors to how often you would expect it to be mutated by chance, given the background mutation rate. Mutations in DNA do not hit every spot equally. Some stretches of the genome are inherently more prone to mutation than others, and this “mutability” matters. Neutral (passenger) mutations tend to occur at sites with higher background mutability, while mutations that recur across many tumors regardless of the local mutability are more likely to be drivers.
5PLOS Computational Biology. Finding driver mutations in cancer: Elucidating the role of background mutational processesAnother key method exploits a natural control built into the genetic code. Some DNA changes alter the protein a gene produces (nonsynonymous mutations), while others do not (synonymous, or “silent,” mutations). If a gene is under positive selection, meaning its mutations help cancer cells, you would expect to see more protein-altering mutations than the silent background would predict. The ratio of nonsynonymous to synonymous mutations serves as a gauge of selective pressure on a gene.
6PubMed Central. Evolutionary dependency of cancer mutations in gene pairs inferred by nonsynonymous-synonymous mutation ratiosMachine learning tools have entered this space as well. Deep-learning platforms can now predict whether a given mutation is likely to be a driver with accuracy above 93% by integrating protein structure, allosteric effects, and population-level mutation frequency data.
7Nucleic Acids Research. DeepAlloDriver: a deep learning-based strategy to predict cancer driver mutations Newer frameworks are even tackling synonymous mutations, which were long assumed to be passengers by definition. It turns out that some silent mutations affect gene regulation, RNA splicing, or protein folding strongly enough to drive cancer, and specialized computational tools are being built to catch them.
8PubMed. MFDSMC: Accurate Identification of Cancer-Driver Synonymous Mutations Using Multiperspective Feature Representation LearningDrivers Hiding in the Non-Coding Genome
For years, the search for drivers focused almost entirely on the roughly 1.5% of the genome that encodes proteins. But the vast non-coding stretches, which include regulatory elements like promoters and enhancers, long non-coding RNAs, and microRNAs, also harbor driver mutations. A landmark analysis of over 2,600 whole cancer genomes identified 705 candidate driver hits across 179 genomic elements, and about one in seven of those hits fell in non-coding regions such as promoters, untranslated regions, and enhancers.
9Nature. Analyses of non-coding somatic drivers in 2,658 cancer whole genomesA dedicated database cataloging these non-coding drivers now lists over 1,100 protein-coding genes and 90 non-coding RNAs with reported driver alterations in their non-coding regions across 32 cancer types.
10PubMed Central. CNCDatabase: a database of non-coding cancer drivers The practical upshot is significant: if a tumor’s driver lies outside the protein-coding region, a gene-panel test that only sequences coding exons will miss it entirely. Whole-genome sequencing catches more, but analyzing the results is far harder because the non-coding genome is enormous and the statistical signal is diluted by the sheer amount of territory.
Passengers Are Not Always Harmless
The classical definition says passengers are neutral. They do not help cancer and they do not hurt it. But a growing body of evidence challenges that tidy picture. Many passenger mutations are, in fact, mildly damaging to the cell. Each one individually is too weak to matter, but collectively they impose a measurable fitness cost. Modeling work suggests that the total burden of deleterious passengers reduces a tumor cell’s fitness by an estimated 46% on average.
11eLife. Most cancers carry a substantial deleterious load due to Hill-Robertson interferenceWhy doesn’t natural selection within the tumor weed out these harmful passengers? The answer involves a phenomenon called Hill-Robertson interference. Because all of a cell’s mutations are inherited together as a package (they are physically linked on the same chromosomes), the cell cannot shed a bad passenger without losing any good drivers on the same chromosome. Tumors with low mutation counts do show signs of purging harmful passengers, but most tumors accumulate mutations faster than selection can remove them.
11eLife. Most cancers carry a substantial deleterious load due to Hill-Robertson interferenceTo cope with this burden, heavily mutated tumors appear to ramp up heat shock pathways, molecular chaperone systems that help proteins fold correctly even when mutations have made them slightly wonky. This buffering lets tumors tolerate their growing load of damaged proteins. Separate modeling work has confirmed that many passengers have largely evaded negative selection and that the collective drag they exert on tumors alters the trajectory of cancer progression in ways that a purely driver-focused view cannot explain.
12PubMed Central. Impact of deleterious passenger mutations on cancer progressionThis is not just an academic curiosity. If passengers genuinely slow tumors down, therapies that increase the mutation rate or magnify the damage done by passengers could push a cancer past the point of viability. Early modeling of this strategy suggests that exacerbating the deleterious effects of existing passengers is more effective than simply piling on new mutations.
12PubMed Central. Impact of deleterious passenger mutations on cancer progressionMini-Drivers and the Gray Zone
Between the textbook categories of strong drivers and neutral passengers sits an uncomfortable middle ground. Mini-drivers are mutations that individually provide only a slight push toward cancer, too weak to qualify as classical drivers by most statistical tests, but strong enough to matter when several accumulate in the same clone. The polygenic mini-driver model proposes that many cancers do not advance through a few dramatic leaps but through the slow, incremental accumulation of many small-effect mutations.
13PubMed Central. Identification of cancer mini-drivers by deciphering selective landscape in the cancer genomeIn colorectal cancer, for instance, researchers have found that multiple accumulated weak mutations can combine to form what amounts to a “polygenic conductor” with enough collective impact to shift cellular function and affect patient prognosis.
14PubMed Central. Impact of mini-driver genes in the prognosis and tumor features of colorectal cancer samples: a novel perspective to support current biomarkers This complicates the binary driver-or-passenger framework, because a mutation’s classification may depend on the company it keeps. A mini-driver that does little on its own might become consequential alongside three others targeting the same pathway.
Why the Distinction Matters for Treatment
The entire premise of targeted cancer therapy rests on identifying driver mutations and designing drugs to block them. If you know a lung tumor is driven by a specific EGFR mutation, you can prescribe a drug that inhibits that mutant protein. The clinical payoff can be substantial compared to broad-spectrum chemotherapy. But there are complications.
First, many tumors harbor more than one driver, and targeting just one often leads to resistance as the cancer falls back on alternative drivers. Drug combination strategies that hit multiple driver pathways simultaneously are an active area of research, but approved drugs do not yet exist for every known driver mutation.
15PubMed Central. Targeted drug combination therapy design based on driver genesSecond, in some cancers, co-occurring driver alterations are rare but clinically tricky. In a large cohort of non-small-cell lung cancer patients, only about 1.7% had two or more driver alterations at the same time. When both alterations were targetable, the most common combination involved EGFR mutations alongside ROS1 rearrangements. Patients with these double-driver tumors who received front-line targeted therapy had a median progression-free survival of only six months, a discouraging number that suggests the drugs interfere with each other or that the biology of double-driver tumors is fundamentally different.
16PubMed Central. Co-Occurring Driver Genomic Alterations in Advanced Non-Small-Cell Lung Cancer (NSCLC): A Retrospective AnalysisThird, catching all the drivers requires adequate testing. When lung cancer patients received comprehensive genomic profiling of their tumor tissue, actionable biomarkers were detected in about two-thirds of cases. Adding a blood-based liquid biopsy on top of thorough tissue testing did not find additional targets. But when tissue testing was limited to a five-gene panel, liquid biopsy detected actionable alterations in an additional 18% of patients, many involving genes like RET, MET, or ERBB2 that the smaller panel simply did not cover.
17JTO Clinical and Research Reports. Complementary Roles for Tissue- and Blood-Based Comprehensive Genomic Profiling for Detection of Actionable Driver Alterations in Advanced NSCLCCollateral Lethality and Exploiting Passenger Deletions
One of the more creative therapeutic ideas to emerge from the driver-passenger framework flips the usual logic. Instead of targeting the driver, you target a vulnerability created by a passenger deletion. When a tumor-suppressor gene is lost through a large chromosomal deletion, the neighboring genes get deleted too, purely because they are physically close on the chromosome. These bystander deletions are passengers in the truest sense: they played no role in cancer promotion.
18Cancer Research. Collateral lethality: A new target for personalized oncologyBut some of those deleted passenger genes had important housekeeping roles. They might have been redundant, meaning the cell has a second copy of a gene that does the same job, so losing one copy does not kill the cell. Normal cells still have both copies. Cancer cells have only one. If you design a drug that inhibits the remaining copy, you selectively kill the cancer cell while leaving normal tissue unharmed. This concept, termed collateral lethality, opens up a large repertoire of potential drug targets because chromosomal deletions in cancer are common and typically span many genes.
19PubMed Central. Collateral Lethality: A new therapeutic strategy in oncologyTumor Evolution and Timing
Not all drivers arrive at the same stage of a tumor’s life. In non-small-cell lung cancer, the canonical drivers like TP53, KRAS, and EGFR mutations are typically clonal, meaning they are found in virtually every cell in the tumor. They were present early, often before the tumor underwent a whole-genome duplication event that doubled its chromosome count. Later-arriving drivers, such as mutations in PIK3CA or chromatin-modifying genes, are more often subclonal, found in only a fraction of tumor cells.
20PubMed Central. Tracking Cancer Evolution through the Disease CourseThis temporal layering produces intratumor heterogeneity: different regions of the same tumor can carry different mutations. Multiregion sequencing of kidney cancers revealed that 63 to 69% of all somatic mutations were not shared across every sampled region of the tumor. Tumor-suppressor genes like SETD2, PTEN, and KDM5C each sustained multiple distinct inactivating mutations in spatially separated parts of the same tumor, a pattern consistent with convergent evolution, where different lineages within the tumor independently hit upon the same functional outcome through different genetic routes.
21PubMed Central. Intratumor heterogeneity and branched evolution revealed by multiregion sequencingFor treatment, this means a single biopsy from one spot in a tumor can miss subclonal drivers lurking elsewhere. The clonal drivers are the reliable targets because they are present everywhere, but subclonal drivers are often the ones responsible for treatment resistance when a targeted therapy kills the dominant clone and a minor subclone with a different driver takes over.
Technical Pitfalls in Classifying Mutations
The tools used to detect and classify somatic mutations are powerful but imperfect. One recurring problem is contamination from germline variants, the inherited DNA differences that have nothing to do with cancer. When a germline variant is mistakenly labeled as a somatic mutation, it distorts the statistical tests used to identify drivers. Germline DNA changes are under different evolutionary pressures than somatic mutations, and mixing them in can either create a false signal of negative selection or mask true positive selection, depending on the direction of the error. Researchers have documented this problem in several prominent datasets, including colorectal cancer and chromophobe renal cell carcinoma cohorts in The Cancer Genome Atlas.
22Cell. Universal Patterns of Selection in Cancer and Somatic TissuesSequencing artifacts, clonal heterogeneity within a sample, and the difficulty of interpreting variants in poorly characterized regions of the genome all add further uncertainty.
23PubMed. Somatic mutations from pathogenesis to forensic medicine: opportunities, challenges, and future directions Ultra-deep sequencing and error-correction methods are improving sensitivity, but no technology currently eliminates all ambiguity. A mutation classified as a passenger in one analysis may be reclassified as a driver when better data or better algorithms become available, and vice versa.
Cross-Species Comparisons as a Filter
One inventive way to separate drivers from passengers exploits evolution across species rather than within a single tumor. If a gene is recurrently gained or lost in both human cancers and cancers from a distantly related species, that shared pattern is unlikely to be coincidence. Comparing copy-number changes in human tumors with those in zebrafish tumors revealed that only about 25% of genes found in human copy-number gains also appeared in zebrafish gains, and only about 22% of losses overlapped.
24PLOS Genetics. Comparative Oncogenomic Analysis of Copy Number Alterations in Human and Zebrafish Tumors Enables Cancer Driver Discovery The overlapping genes are enriched for true drivers, because a passenger alteration in a human cancer has no reason to also show up in a fish cancer. The non-overlapping majority is likely passengers or species-specific biology. This kind of cross-species filtering provides an independent check on computational predictions and has helped prioritize candidate driver genes for experimental follow-up.
Epigenetic Drivers
The driver-passenger distinction does not stop at DNA sequence changes. Epigenetic alterations, chemical modifications to DNA or its packaging proteins that change gene activity without changing the underlying sequence, can also drive cancer. Abnormal methylation that silences a tumor-suppressor gene, for example, can have the same functional effect as a mutation that breaks the gene. These epigenetic events affect multiple genes and pathways in a non-random fashion and can predispose cells to accumulate further genetic damage as a tumor develops.
25PubMed. Epigenetic drivers and genetic passengers on the road to cancerEpigenetic passengers exist too: methylation changes that drift along with tumor expansion without contributing to it. Sorting the two is even harder than sorting genetic drivers from passengers, because epigenetic marks are more context-dependent and more reversible than DNA mutations. This reversibility, however, also makes epigenetic drivers attractive therapeutic targets. Drugs that restore normal methylation patterns or modify histone marks are already approved for certain blood cancers, and the logic behind their use is essentially the same as targeting a genetic driver: find the change that is helping the cancer, and undo it.
Driver Mutations in Healthy Tissue
Perhaps the most unsettling finding to emerge in recent years is that driver mutations are not unique to cancer. They are widespread in normal, healthy-looking tissues. Skin, esophagus, blood, liver, and other organs all accumulate clones carrying known cancer-driver mutations as people age. These clones expand over time, sometimes occupying a substantial fraction of the tissue, without ever forming a tumor.
26PubMed Central. Somatic Mutations in Normal Tissues: Calm before the StormThis observation raises uncomfortable questions. If a driver mutation is defined by its ability to promote cancer, what do you call the same mutation sitting quietly in tissue that shows no signs of disease? The answer seems to be that context matters enormously. A driver mutation may need additional hits, the right tissue microenvironment, inflammation, or the failure of immune surveillance before it can actually drive anything. External factors like chronic inflammation appear to enhance the competitive fitness of mutant clones, tipping the balance from peaceful coexistence to malignant transformation. Understanding this pre-cancer landscape may eventually enable interception strategies that prevent tumors before they start, rather than treating them after they appear.