Home » Blog » "Science" » Medicine » Oncology » From Paracetamol and Alzheimer’s Loops to a Minimal Viable Cancer Loop (MVCL)

From Paracetamol and Alzheimer’s Loops to a Minimal Viable Cancer Loop (MVCL)

Part 3 of a trilogy

This post is essentially Part 3 in a trilogy:

  1. Paracetamol:
    Paracetamol’s braided mechanism – how we finally closed a 138-year pharmacology puzzle
    https://hmwh.se/blog/2025/11/24/paracetamols-braided-mechanism-how-we-finally-closed-a-138-year-pharmacology-puzzle/

  2. Alzheimer’s / Dementia:
    From a walk home to a testable Alzheimer’s theory
    https://hmwh.se/blog/2025/08/14/from-a-walk-home-to-a-testable-alzheimers-theory/

    From a testable Alzheimer’s model to an ion-terrain map of dementia
    https://hmwh.se/blog/2025/11/13/from-a-testable-alzheimers-model-to-an-ion-terrain-map-of-dementia/

  3. This post (Cancer):
    How we took the same AI-assisted “theory template” and forced cancer into a Minimal Viable Cancer Loop (MVCL) with a pan-cancer playbook on top.

1. Things I’ve been poking at with AI

Very short recap of the first two parts:

  • For paracetamol, we ended up with a braided, multi-lane mechanism with explicit equations and falsifiers – not “one magic receptor”, but a four-lane central mechanism that only makes sense when you track state, timing and context carefully.

  • For Alzheimer’s, a lithium thought on a walk home turned into DISSAD (a DMN + Li⁺ + tau/GSK3β model) and DISSAD+, an “ion-terrain” trunk-and-fork map of dementias where Alzheimer’s is just Config-1 on a stressed DMN backbone.

Underneath both projects, the AI kept pushing the structure into the same pattern:

  1. A minimal loop or trunk that can actually run.

  2. Around it, modules that each ship:

    • a short causal chain (≤ ~7 steps),

    • invariants vs context,

    • biomarkers / readouts,

    • control levers,

    • and 1–3 hard falsifiers.

  3. Everything written down tightly enough that other people can try to kill it.

So the obvious next step was: what happens if you give cancer the same treatment?

2. Giving the template to cancer

Cancer is the canonical “too big to think about in one go” disease:

  • countless hallmarks,

  • every pathway someone’s favourite,

  • huge literature, very few compact architectures.

The AI-assisted challenge I set up was:

Take the paracetamol + DISSAD template and see if there is a minimal loop-of-loops that all lethal tumours must implement, regardless of tissue, driver mutations or drug history.

The constraints:

  • Minimal loop first, details second.

  • Every component must earn its place via:

    • mechanistic plausibility,

    • near-term testability,

    • and a clear decision consequence if it’s falsified.

  • Everything should be expressible in CSV + diagrams, not only prose.

That process gave us the Minimal Viable Cancer Loop (MVCL).

3. The Minimal Viable Cancer Loop: Seed → Switch → Sink → Spread

In MVCL v1.2, lethal cancer is abstracted as four interlocking phases:

  • Seed – processes that generate variation or bias the host terrain
    (somatic mutations, telomere crisis, aging/CHIP, chronic inflammation, etc.)

  • Switch – circuitry and state transitions that permit survival under stress
    (oncogene/TSG rewiring, checkpoint erosion, lineage switching, apoptosis gating…)

  • Sink – exploitable dependencies and permissive niches
    (replication-stress tolerance, immunosuppressive TME, metabolic crutches…)

  • Spread – dissemination and evolutionary propagation
    (EMT/pre-metastatic niche, clonal selection, resistance reseeding…).

The claim is modest but sharp:

If a system cannot implement some version of
Seed → Switch → Sink → Spread,
it cannot sustain a lethal, evolving cancer in a human.

Destroy any one leg hard enough and the loop collapses.

4. Thirteen modules wrapped around the loop

Around that loop sit 13 core “backbone” modules (silos). Each module is a small theory box with:

  • a short causal chain,

  • invariants vs context,

  • biomarkers/readouts,

  • control levers,

  • and a key falsifier.

Examples:

  • Somatic mutation & clonal evolution

  • CIN, aneuploidy, WGD & chromothripsis

  • DDR & replication stress

  • Oncogene/TSG circuitry & non-oncogene dependencies

  • Structural variation & ecDNA

  • Epigenetic reprogramming & lineage plasticity

  • Metabolic rewiring

  • Tumour microenvironment (CAF/ECM/hypoxia/angiogenesis/acidosis)

  • Immune evasion / tumour immune microenvironment

  • Cell death & senescence

  • Metastasis / EMT / pre-metastatic niche

  • Therapy resistance & evolution

  • Aging / CHIP overlays

Instead of writing one giant, linear review, we force each module to declare a handshake via an edges_*.csv:

  • which other modules it talks to (source_module, target_module),

  • by what mechanism and with what effect (stimulate / inhibit / reseed),

  • in which context (acute DDR vs chronic WGD, primary vs metastatic niche, etc.),

  • how we’d see that edge (readouts),

  • and how we might pull on it (control levers).

The result is a wiring diagram: a loop-of-loops that you can eventually simulate and design studies against, not just a wall of text.

5. The four pillars: a universal pan-cancer playbook

Once the MVCL backbone was stable, I asked a more practical question:

“If you had to walk into Tumour Board with one strategic playbook for aggressive tumours,
what bottlenecks would you always try to hit?”

That’s where the four pillars fell out, as written up in the MVCL Universal Playbook v1.

Pillar 1 – Deny the Sanctuary (Niche & Immune Shield)

Objective: collapse the protective niche / immune-evading environment.

  • Normalize TME (reduce hypoxia, improve vessels, buffer acidosis).

  • Block immunosuppressive metabolites (lactate/MCT, adenosine/A2A).

  • Address systemic shields (aging/CHIP, inflammaging, senescent SASP).

Monitoring: pH/lactate, A2A expression, vessel maturity, CHIP VAF, SASP signatures.
Levers: LOX/FAK/YAP-TAZ axis, A2A antagonists, MCT inhibitors, senolytics, etc.

Pillar 2 – Crash the Replication Engine (High-Stress Division)

Objective: exploit dependence on replication stress and CIN.

  • Stratify by RS/CIN signatures.

  • Deploy ATR/CHK1/WEE1 and related RS-support levers.

  • Combine with IO when cGAS–STING is still functionally intact.

Monitoring: γH2AX, pCHK1/2, RS signatures, ecDNA/WGD scores.
Levers: ATR/CHK1/WEE1 inhibitors, nucleotide-pool disruptors.

Pillar 3 – Force the Death Decision (Apoptosis Gate)

Objective: push malignant cells past a real death threshold.

  • Use BH3 profiling to gauge how “primed” cells are.

  • Combine BH3 mimetics (BCL-2, MCL-1, etc.) with RS/metabolic stresses.

Monitoring: BH3 priming profiles, BCL-2/MCL-1, glutamine flux.
Levers: BH3 mimetics, glutaminase/IDH inhibitors and other apoptosis sensitisers.

Pillar 4 – Block the Escape Route (Plasticity / Tolerance Engine)

Objective: shut down lineage switches and drug-tolerant persisters.

  • Short, timed pulses of epigenetic modifiers (EZH2, HDAC, CDK7/9).

  • Intercept EV/ncRNA signals that spread plasticity and pre-metastatic niche programmes.

Monitoring: DTP ATAC/RNA signatures, EV integrins/miRNAs, single-cell plasticity metrics.
Levers: epigenetic reprogrammers, EV-pathway blockers.

The Playbook also sketches:

  • Sequencing heuristics – profile → create a window of vulnerability (Pillars 1–3) → apply escape blockade (Pillar 4) → then move into maintenance.

  • A monitoring matrix with key readouts and rough thresholds.

  • Example “regimen archetypes” that show how timing might look in practice.

6. Edge cases, caveats and refinements

The MVCL Updates v1.2 + Edge Cases document tracks where this architecture needs nuance. A few highlights:

  • Immune extremes:

    • In profound immunodeficiency (post-transplant, uncontrolled HIV), IO-heavy plays may underperform; you lean harder on niche disruption + RS/apoptosis (Pillars 1–3).

    • In autoimmune-prone patients, aggressive IO stacking is risky; more weight shifts to RS and TME control.

  • Tumours with low RS / indolent proliferation:
    Not every tumour is RS-high. When RS signatures are low, Pillar 2 becomes a lighter touch; TME, metabolism and plasticity levers take more of the load.

  • Highly immunogenic, low-CIN tumours (e.g. some MSI-H / dMMR):
    IO + TME normalization may already be powerful; RS levers are used more cautiously.

  • Timing refinements:

    • Pillar 1: make explicit the vascular normalization window (use imaging like DCE-MRI to time access).

    • Pillar 2: stagger ATR/CHK1/WEE1 and heavy cytotoxics to avoid overlapping toxic peaks.

    • Pillar 3: re-profile BH3 priming after 1–2 cycles; it can shift quickly.

    • Pillar 4: epi pulses must be long enough to actually remodel chromatin; ultra-short micro-doses risk being biologically irrelevant.

In other words: the MVCL + Playbook is an architecture, not a magic regimen. Context still rules.

7. The MVCL OSF package

Everything described above lives in an open OSF project:

MVCL – Minimal Viable Cancer Loop
OSF: https://osf.io/gh6nm/

Key components:

  • Integrated Minimal Viable Cancer Loop (MVCL) Framework.pdf
    High-level conceptual framework and the Seed → Switch → Sink → Spread map.

  • Minimal viable cancer loop.pdf
    Visual loop + compact narrative explainer.

  • Cancer Synthesis — Backbone Map & Workflow.pdf
    How the loop, modules and workflows fit together in one page.

  • Cancer Synthesis — External Explainer (for Scholar GPT_consensus).pdf
    A more accessible text explainer for external readers / reviewers.

  • Backbone sheets for all silos (+ “Backbone Sheets — All Silos (master).pdf”)
    One detailed “backbone sheet” per module.

  • edges_v1_2.csv
    The machine-readable edge list (module-to-module interactions, contexts, readouts, levers).

  • Mvcl Universal Playbook V1.pdf
    The four-pillar pan-cancer strategy, with timing heuristics, monitoring matrix and lever classes.

  • Mvcl Updates V1 2 Plus Edge Cases.pdf
    Edge cases, guardrails and refinements to the pillars and monitoring.

  • Uncertainties & Falsifiers — All Drafted Silos (v1).pdf
    Per-module uncertainties and falsifier lists – the “how to kill this” appendix.

Together, these form a minimal but testable cancer OS built with the same internal API as:

  • the braided paracetamol model, and

  • the DISSAD / DISSAD+ ion-terrain for dementia.