One Landscape, Many Inputs: Cancer Prevention as Cofactor-and-Gene Route-Competition

Paper 8E · Pødenphant Lund, T. (2026) · Hypothesis paper · Live on Zenodo

Three bodies of prevention evidence are usually told as three separate disappointments: micronutrient-supplement trials that "fail to replicate" (folate, selenium, β-carotene, vitamin E protect in some populations and do nothing, or harm, in others); gene–diet-interaction studies that accumulate as a case-by-case catalogue with no general rule; and neurodevelopmental risk genes that turn out to be cancer genes, even though carrying more variation in cancer-driver genes can track with less clinical cancer. This paper proposes they are one signature, not three failures. If carcinogenesis is a commit-with-hysteresis race organised by a single order parameter, the tissue's baseline trace-state, then the same input tilts toward repair when baseline trace is low and toward an already-committed clone when it is high, so it protects some and harms others. It is a hypothesis and an invitation to test it, with two discriminating predictions stated with their falsifiers. It is not medical advice and recommends nothing at the individual level.

DOI (concept)	10.5281/zenodo.21470791
Type	Hypothesis paper (framework synthesis of published evidence, with falsification criteria)
Order parameter	Baseline trace-state (tissue damage / early-clone burden)
Worked exemplar	One-carbon / folate–MTHFR cycle; alcohol as its negative cofactor
Author	Tomas Pødenphant Lund [ORCID]

The construct: a commit-with-hysteresis race and its order parameter

Carcinogenesis is modelled as a competition between two route-classes in a tissue: genome-maintenance and differentiation on one side, damage-accumulation and proliferation on the other. The race is a commit-with-hysteresis process: once the proliferative route has been committed, reversal is expensive, so the system carries the trace of its own history. The framework's single organising variable is the baseline trace-state, the amount of damage and early clonal change already present in a given tissue. This is the order parameter that makes the whole account non-monotone: an input does not have a fixed sign of effect; its sign is read off the trace-state of the tissue it acts on. At low trace, an input that feeds a rate-limiting maintenance route tilts the race toward repair (protection); at high trace, the same input also supplies a clone that is already winning (promotion). Same input, opposite effect, because the landscape underneath it has moved.

The cross-system signature: one shape across unrelated inputs

Averaging a randomised trial over a mixture of low-trace and high-trace participants cancels a real subgroup effect toward the null, which is the framework's reading of why the supplement literature looks inconsistent. Read by subgroup rather than by headline, the low-trace-protect / high-trace-harm signature recurs across nine of fifteen surveyed, chemically unrelated supplement–cancer systems, with two informative non-fits (vitamin C and vitamin D) that bound the claim to inputs feeding a rate-limiting route in an active trace-accumulating process. The recurrence across chemically unrelated inputs is the load-bearing observation: it is what distinguishes one shared signature from a set of unrelated disappointments, and the two non-fits are treated as a prediction the framework makes, not an exception waved away.

Alcohol as a negative cofactor

The cleanest handle is an input that pushes the landscape the other way. Alcohol depletes folate, so the framework predicts it should carry the mirror signature: harm concentrated in the already-low-folate stratum. It does, and this offers a testable resolution of a long-standing discordance. Observational studies link alcohol to breast cancer clearly, yet Mendelian-randomization estimates of the causal effect are close to null. Both can hold: the genetic-causal estimator averages over the whole population and washes out an effect that lives only in the folate-low, the same dilution that flattens the supplement trials. An effect that is real but landscape-dependent is invisible to a method that sums over the landscape.

The gene route, and the paradox it dissolves

The same landscape logic extends to genes. Tumour-suppressor and chromatin/growth-signalling genes set the steepness of the commit-landscape, and a subset of them overlap the high-confidence neurodevelopmental risk genes. The framework's reading is that a risk gene shifts the commit-landscape but does not determine the commit, which stays gated by the cofactor environment. The strongest variants push so far that the race commits almost regardless of environment (hence very high penetrance); common weak variants push only sub-threshold, so at population scale their effect dilutes and can invert. The apparent paradox of more cancer-gene variation with less clinical cancer therefore dissolves into a prediction rather than standing as an anomaly: carriage moves the terrain, the environment gates whether the edge is crossed. The paper keeps this as a claim about landscape, not diagnosis, and flags that the direct gene–cancer association is clearest for colorectal, not breast.

Predictions and falsifiers

The predictions are separated by how much they discriminate the framework from the nearby accounts (hormesis, baseline-status, multi-stage latency, generic gene–environment interaction). Two discriminate and carry the paper; two are consistent with it but shared with rivals and are marked as such.

• C★ (discriminating) — the active form widens the window. Because the protect-to-promote flip is set by a cofactor's rate-limiting activation step, the post-bottleneck active form should reach the promotion stratum only at a higher dose. Concretely: in MTHFR C677T carriers, 5-methyltetrahydrofolate should have a wider cancer-safe window than folic acid (whose under-conversion accumulates unmetabolised folic acid). Neither hormesis nor baseline-status says anything about chemical form. Falsifier: C677T carriers show the same or a narrower window with the active form, or no form difference, on a cancer-relevant endpoint. Status: untested for any cancer endpoint.
• G★ (discriminating) — directional gene × cofactor. Not the generic "gene effect depends on diet" (every gene–environment framework predicts that), but a directional, gene-set-specific claim: carriers of the curated commit-landscape set (the autism/ADHD tumour-suppressor and chromatin/growth-signalling overlap, e.g. PTEN, NF1, TSC1/2, CHD8, ARID1B) should show a folate/alcohol interaction of the same sign as the cofactor systems, and of larger slope than a matched control set of neurodevelopmental genes that are not cancer drivers. Falsifier: the commit-landscape set shows no steeper cofactor interaction than the control set, or an interaction of opposite sign. Test: a biobank gene-set analysis with sets fixed in advance (Cancer Gene Census intersected with SFARI-type genes; control set matched on length, constraint, expression), primary contrast the between-set difference in the cofactor-interaction coefficient, with permutation control. Status: untested; the prediction that unifies the gene and cofactor threads.
• P★ (supporting) — hysteresis in time. The protect-to-promote transition is history-dependent (lagged, or persistence-after-removal harm). Multi-stage latency predicts the same aggregate time-course, so this discriminates only at the individual level. Sharpened falsifier: among individuals matched on current trace-state and cumulative dose, early-heavy versus late-heavy exposure histories show no difference in future risk (a pre-specified path contrast in a multi-state model).
• D★ (supporting) — the signature generalises, named in advance. The framework predicts the protect-low / harm-high shape for methionine supplementation and hepatocellular carcinoma, and for copper and colorectal cancer, and predicts its absence for a pre-specified negative control (a nutrient not rate-limiting for any committed route, e.g. vitamin C at a second site). Falsifier: a named positive system shows a uniform trace-independent effect, or the negative control shows the signature.

Bounds and prior art

Most of the biology is borrowed and heavily studied: one-carbon metabolism and its coupling to methylation and nucleotide supply, the folate-timing / dual-modulator hypothesis, gene–diet interaction, the metabolic-control-analysis account of multi-factor prevention, and the autism–cancer gene overlap are each mature literatures. The paper does not claim to have discovered that folate's effect is non-monotone, that baseline status modifies supplement effects, or that autism and cancer share genes. The two nearest existing readings, hormesis and the baseline-status "supplement the deficient, not the replete" account, already cover part of the ground. What the lens adds beyond them, and what the paper stakes itself on, is narrower: the temporal hysteresis signature, the extension of the landscape logic to the gene route, and the two discriminating predictions above. The paper is explicit that a smooth, path-independent, cofactor-insensitive result would return the idea to re-description.

What this paper is, and is not

What it is: a hypothesis and an invitation to test it; a single order-parameter reading that unifies three anomalies (the supplement-trial mess, the gene–diet catalogue, the alcohol observational-vs-causal discordance); a survey placing the signature across nine of fifteen chemically unrelated systems with two bounding non-fits; and four predictions with falsifiers, the two discriminating ones runnable, in first instance, as re-analyses and biobank gene-set analyses on existing data.

What it is not: it is not medical advice, and it recommends no supplement, dose, or product to anyone. The active-form point (C★) is a research prediction, explicitly untested for cancer, not a recommendation to take 5-MTHF. The gene claim is about landscape, not diagnosis, and is never a statement about people with a diagnosis. The mathematics of the race reading is a deliberate simplification of a multi-step biology.

Connections to other papers in the series

• Paper 1 (Friction Theory) — the substrate-universal framework of competing routes under commit-pressure.
• Paper 10 (Race architecture) — the substrate-general race vocabulary (parallel candidates, bounded resources, irreversible commit) this paper applies to tissue.
• Paper 8 (Pressure, Hysteresis, and Experience) — the clinical foundation and the additive-only hysteresis asymmetry the prevention reading rests on.
• Paper 8B (Compound Race Pathology) — the multi-scale combination logic; the cofactor-and-gene reading here is its prevention-side counterpart.
• Paper 8D (Treating the Base) — the sub-threshold cofactor-support companion; 8D matches support to a measured profile below the diagnostic threshold, 8E asks when the same cofactor protects versus harms.

Read the paper

The full paper is on Zenodo (concept DOI 10.5281/zenodo.21470791):

Read on Zenodo → · Plain English version · Dansk version

This page summarises a hypothesis paper. It is not medical advice and does not tell any individual what to do. The biological claims are predictions to be tested, not proven facts, and the active-form point is explicitly untested for cancer. Any decision about supplements, testing, or treatment belongs with a qualified clinician.