Clonal Interference
Summary
Clonal interference is the competition between two or more expanding clones that carry different beneficial mutations, such that no single clone can complete a full clonal-sweep before the next advantaged clone arises. The result is incomplete, overlapping expansions rather than sequential, complete sweeps — the characteristic pattern of branching-evolution in solid tumors. Clonal interference arises when the waiting time for the next driver mutation is shorter than the time required for the current driver-bearing clone to sweep to fixation, a condition that becomes increasingly likely as tumor size grows. Greaves & Maley (2012) identified clonal interference as the dominant evolutionary dynamic in established solid tumors, driven by the combination of large population sizes (N ~ 10⁸–10¹¹ cells), modest driver fitness advantages (~0.4% on average), and ongoing mutation supply. It is the null model against which the clean clonal-sweep — Nowell’s (1976) original picture — must be tested.
The Transition from Sweeps to Interference
The evolutionary mode of a tumor is determined by the relationship between two timescales (clonal-sweep §Timescale Competition):
- τ_k — the waiting time for the next driver mutation to appear in a tumor of size N with driver mutation rate u per cell division
- sweep time — the time for a clone with selective advantage s to expand from a single cell to fixation in a population of size N
From the Bozic et al. (2010) branching process model:
where T is the cell division time, k is the number of existing drivers, and s is the selective advantage per driver.
When τ_k > sweep_time, each driver sweeps to fixation before the next arises — clean sequential sweeps, Nowell’s regime. This is the condition in small populations (early tumors, N ~ 10³–10⁵).
When τ_k ≤ sweep_time, a new driver mutation appears before the previous clone has completed its sweep. The two clones compete for resources and space — clonal interference. Neither achieves full fixation. This is the condition in large populations (established tumors, N ~ 10⁸–10¹¹).
The transition is driven primarily by tumor size: as the population grows, the target size for new driver mutations increases (more cells = more opportunities), so the waiting time shrinks, while the sweep time grows (a clone must traverse a larger population to reach fixation). The condition τ_k > sweep_time that guarantees clean sweeps becomes harder to satisfy, and eventually impossible (clonal-sweep §Timescale Competition).
Why Clonal Interference Dominates in Solid Tumors
Greaves & Maley (2012) identified three factors that make clonal interference the predominant dynamic in established cancers:
1. Large population sizes. A 1 cm³ tumor contains ~10⁹ cells. A clinically detected tumor (~5 cm³) contains ~5 × 10⁹ cells. The sweep time for a single clone to traverse this population — even with a strong selective advantage — is measured in years. Meanwhile, the waiting time for a new driver mutation somewhere in this vast population is measured in months or weeks.
2. Modest selective advantages. Bozic et al. (2010) estimated driver fitness advantages of ~0.4% on average from glioblastoma and pancreatic cancer data (as cited by Greaves & Maley, 2012). A 0.4% advantage means that a driver-bearing clone produces 1.004 descendants for every 1.000 produced by a non-driver-bearing competitor. This is a weak advantage — the sweep is slow, giving competitors ample time to arise.
3. Ongoing mutation supply. Mutational processes are not confined to early tumorigenesis. APOBEC mutagenesis, clock-like signatures, and therapy-induced mutagenesis continue to generate new mutations — including new driver candidates — throughout the tumor’s lifetime. The ongoing supply of variation means that even if one clone were to approach fixation, new competitors continually appear.
The result is that clean, complete selective sweeps are the exception rather than the rule in clinically detected solid tumors. Clonal interference — multiple clones with different driver mutations expanding simultaneously, competing for resources and space — is the dominant dynamic (Greaves & Maley, 2012; clonal-evolution).
Consequences
Branching Architecture
Clonal interference produces branching-evolution: multiple subclones with different driver mutations coexist as siblings from a common ancestor, with no single clone achieving dominance. This is the most common phylogenetic architecture observed in multi-region sequencing studies (Turajlic et al., 2019; intratumor-heterogeneity §4).
Incomplete and Overlapping Expansions
Under clonal interference, clonal-expansions are “often incomplete and overlapping rather than sequential and complete” (clonal-expansion). A clone may expand to occupy 30% of the tumor, then stall — not because it lost its advantage, but because a competitor with a different advantage expanded into the remaining territory. The final clonal architecture is a mosaic of overlapping, incomplete expansions.
Detection Challenges
Clonal interference complicates the detection of selection from genomic data:
- The 1/f² VAF distribution test for neutral-evolution can fail to detect weak selection under clonal interference — multiple competing clones with similar fitness advantages produce a VAF distribution that deviates from the neutral expectation less dramatically than a single strong sweep would (molecular-clock §Selection Detection)
- The crossing rule from multi-region sequencing (crossing-rule) is the primary tool for identifying clonal interference: spatially variable clone abundances that cross between regions reveal sibling relationships, not dominance
- Single-biopsy data under clonal interference can misrepresent the tumor: a biopsy that captures the territory of one competing clone shows a monoclonal picture; a biopsy from the territory of another clone shows a different monoclonal picture. Only multi-region data reveals the underlying interference pattern
ITH Maintenance
Clonal interference actively maintains ITH: because no single clone can sweep to fixation, diversity is continuously replenished by new driver-bearing clones competing with existing ones. This is the ITH regime (moderate SMF, 0.20–0.60) predicted by the compression-entrenchment hypothesis to be associated with better outcomes — the tumor remains in a transitional, un-entrenched state because clonal interference prevents any single clone from achieving complete compression (intratumor-heterogeneity §5).
Clonal Interference in the Formal Models
The Cancer Evolution Olog
In the cancer-evolution-olog, clonal interference is formalized as the failure of commutativity in the weak-selection regime. The arrow sweep from DriverMutation → ClonalSweep (a driver mutation causes a sweep) is defined only on the subcategory where s ≫ 1/N (strong selection). In the clonal interference regime (s ~ 1/N), multiple DriverMutation events compete and no single ClonalSweep is defined — the commutativity condition that guarantees deterministic, sequential sweeps fails (cancer-evolution-olog §Challenge).
The Bozic-Nowak Timescale Condition
The transition from sweeps to interference is formalized through the Bozic-Nowak timescale separation (clonal-sweep §Timescale Competition). Clean sweeps occur when:
Clonal interference occurs when:
This is the mathematical formalization of the Greaves & Maley (2012) insight that “clonal evolution is not always successive selective sweeps.”
Cross-Domain Functors
In the cross-domain-functors analysis, clonal interference marks a key breakdown of functorial commutativity:
- Functor G (Compression → Cancer): A ClonalSweep is a Discovery (a compression breakthrough) iff τ_k > sweep_time. If a new driver appears before the sweep completes, the sweep is incomplete — the discovery is “blurred” across multiple clones. The functor holds in the strong-selection, clean-sweep regime; it breaks in the clonal interference regime (cross-domain-functors §G-CC5).
- Functor F (Ecology → Cancer): The ecological analogue of clonal interference — multiple invading species with different adaptations competing for the same niche — maps cleanly onto cancer. This is one of the few cross-domain mappings that holds across both the sweep and interference regimes: clonal interference in cancer is structurally identical to competitive exclusion failure in ecology.
Clonal Interference and Therapy
Clonal interference has direct implications for therapy:
Polyclonal resistance. Under clonal interference, multiple clones with different driver mutations coexist. When therapy is applied, each clone may have a different resistance mechanism — and because the clones are already present at detectable frequencies (not below the detection floor), the resistance is polyclonal from the start. This is worse than monoclonal resistance (one clone, one mechanism) because multiple resistance mechanisms must be targeted simultaneously (Turajlic et al., 2019).
The bottleneck paradox. Therapy imposes a population-bottleneck that can temporarily eliminate clonal interference by killing most clones. But the surviving clones, released from competition with the eliminated clones, expand rapidly — and the interference dynamic re-emerges as mutation generates new competitors. The post-therapy relapse is often a return to clonal interference, not a monoclonal escape (Miething, 2019; population-bottleneck).
Ecological therapy. Greaves & Maley (2012) proposed ecological therapy as an alternative to maximum-tolerated-dose cytotoxic therapy: instead of attempting to eliminate all clones (which selects for resistant ones), manage the competitive dynamics to prevent any single clone from dominating. This is the therapeutic analogue of clonal interference — using competition between clones as a control strategy rather than a problem to be overcome (clonal-evolution).
Limitations
Clonal interference is difficult to observe directly. It requires multi-region or longitudinal sequencing at sufficient depth to resolve clone frequencies, and the statistical methods for distinguishing true clonal interference from neutral coexistence (clones with no fitness differences coexisting by drift) are underdeveloped. Many tumors classified as showing branching architecture may in fact be neutral rather than subject to clonal interference — the alternative mechanisms produce similar phylogenetic patterns from different evolutionary dynamics.
The fitness effects of competing clones are usually unknown. Clonal interference is defined by competition between clones with different beneficial mutations. But in practice, the fitness effects of the private mutations distinguishing coexisting clones are rarely known — are they drivers (clonal interference) or passengers (neutral coexistence)? The distinction matters for therapy (interfering clones each have distinct vulnerabilities; neutral clones share the same driver landscape) but is observationally ambiguous.
The transition from sweeps to interference is model-dependent. The Bozic-Nowak τ_k condition assumes a specific branching process parameterization. Real tumors have spatial structure (which slows sweeps, favoring interference), variable driver fitness effects (some drivers are strong enough to sweep despite interference), and microenvironmental heterogeneity (which can shelter clones from competition, reducing interference). The clean transition predicted by the model is a useful idealization, not a precise description of any individual tumor.