Note (2026-08-28): the Metastatic_* entries named in this project were folded into their histologic parent entries per design decisions §3a (Metastatic_NSCLC → Non-Small Cell Lung Cancer, Metastatic_Colorectal_Cancer → Colon Adenocarcinoma, Metastatic_Pancreatic_Adenocarcinoma → Pancreatic Ductal Adenocarcinoma, Metastatic_Renal_Cell_Carcinoma → Renal Cell Carcinoma). Historical tables below retain the old names.

Tumor Microenvironment Modeling

Overview

Explore how dismech's structured mechanism models can serve as a knowledge substrate for computational tumor microenvironment (TME) modeling, especially multiscale agent-based simulators, reinforcement-learning-guided therapy optimization, and cancer digital twin frameworks.

The core idea is straightforward: dismech already encodes disease mechanisms as cell types, biological processes, and downstream causal effects. Those graphs are close to the rule systems that agent-based tumor simulators need, but they are currently optimized for human-readable disease explanation rather than for machine-actionable simulation. This project asks what is missing to bridge that gap.

STATUS

• [ ] Define a TME completeness rubric for existing cancer entries: immune, stromal, vascular, metabolic, spatial, macroenvironment
• [ ] Review the colorectal cancer entries against that rubric and document concrete missing mechanism actors
• [ ] Design a reusable tumor_ecosystem module analogous to fibrotic_response and immune_checkpoint_blockade
• [ ] Decide whether claim extraction in issue #1100 should be the canonical interchange layer for simulation exports
• [ ] Specify a minimal simulator-facing rule schema beyond raw triples
• [ ] Prototype export of dismech mechanism graphs into a neutral interaction_rules format
• [ ] Map existing immune_checkpoint_blockade nodes into the broader tumor ecosystem module
• [ ] Identify high-value CRC additions: CAFs, TAMs, MDSCs, endothelial cells, hypoxia, angiogenesis, metabolic suppression, spatial niches
• [ ] Select 2-3 non-CRC cancer entries for the next TME-enrichment pilot
• [ ] Draft a small PhysiCell-oriented proof of concept showing how one curated mechanism chain becomes executable cell-cell rules

Scientific Thesis

Cancer outcome depends on the interaction of four layers:

1. Tumor cells
2. The immune microenvironment
3. The stromal microenvironment
4. The systemic macroenvironment

The working thesis for this project is that tumor-targeted monotherapy fails because genomic and epigenomic instability generate resistance heterogeneity faster than tolerable tumor-cell-only drug combinations can suppress it. Therapeutic strategy therefore has to shift from "kill the clone" to "re-engineer the ecosystem."

Three motivating principles:

1. Tumor heterogeneity creates heterogeneous treatment response and resistance that cannot be managed with tumor-targeted combinations alone.
2. Drugs aimed at tumor cells also reshape immune, stromal, and systemic environments, sometimes in pro-tumor and sometimes in anti-tumor directions.
3. Better efficacy and tolerability should come from combination strategies that jointly target tumor cell state, the TME, and the macroenvironment to push the whole system toward a globally anti-tumor equilibrium.

Why dismech Fits This Problem

dismech is already closer to simulator-ready knowledge than a literature review or pathway diagram:

• Pathophysiology entries are structured causal chains rather than prose blobs.
• Mechanism modules already capture reusable cross-cell programs.
• Treatments can already point back to mechanism nodes through target_mechanisms.
• The planned claim-extraction work in issue #1100 should make these chains serializable as machine-queryable interaction statements.

In other words, dismech can potentially become the mechanistic prior layer that many TME simulators currently hand-author.

Existing dismech Anchors

CRC is therefore the natural pilot. The KB already has subtype-level disease coverage, at least one explicit tumor-immune module, and a metastatic entry that starts to gesture toward stromal and vascular ecology.

Active Computational Landscape

Gap Analysis for Current CRC Coverage

The existing colorectal entries are a strong start, but they are not yet rich enough to drive TME simulation without substantial hand engineering.

Immediate CRC priorities

1. Add explicit immune suppressor and stromal effector cell types to the colorectal entries: CAFs, TAMs, MDSCs, Tregs, endothelial cells, pericytes.
2. Add missing processes that are central to simulator behavior: angiogenesis, hypoxia response, extracellular matrix remodeling, leukocyte trafficking, metabolic competition, and immune exclusion.
3. Represent organ-specific metastatic ecology in metastatic CRC, especially the liver niche, rather than treating metastasis as a tumor-cell-only phenotype.
4. Capture subtype-specific TME distinctions instead of one generic CRC environment. MSI-high, BRAF-mutant, HER2-positive, and metastatic CRC should not share the same default ecosystem logic.

Proposed tumor_ecosystem Module

This should be a reusable mechanism module in kb/modules/ using the same schema as existing disorder entries.

Design principles

1. Keep the module generic enough to reuse across solid tumors.
2. Let conforming entries substitute organ- and tumor-specific cell types.
3. Reuse existing module nodes where possible instead of duplicating them.
4. Separate minimal conserved logic from disease-specific specialization.

Candidate conserved nodes

Relationship to existing modules

• immune_checkpoint_blockade#Adaptive Immune Resistance should remain a reusable conformance target inside the broader tumor ecosystem model.
• fibrotic_response is a useful design precedent for stromal remodeling and myofibroblast-like state transitions.
• A future angiogenesis_inhibition module would likely plug into this same ecosystem architecture.

Export Format for Agent-Based Models and Digital Twins

Raw triples are necessary but not sufficient.

The claim-extraction roadmap in issue #1100 is important because it should produce machine-queryable statements such as:

(cell_type_A, process_X, cell_type_B)

However, a simulator needs more than that. It also needs direction, sign, conditions, spatial scope, and action semantics.

Recommended layered export

1. Claim triples Minimal extraction layer for indexing and search.
2. Rule objects Simulator-neutral mechanism objects with enough fields to execute or compile.
3. Backend adapters Translators from the neutral rule objects into PhysiCell, TumorTwin, or other framework-specific inputs.

Minimal simulator-neutral rule schema

{
  "disease": "MSI_High_Colorectal_Cancer",
  "module_node": "immune_checkpoint_blockade#Adaptive Immune Resistance",
  "source_cell_type": "tumor cell",
  "signal_or_process": "PD-L1 upregulation",
  "target_cell_type": "CD8-positive, alpha-beta T cell",
  "target_process": "T cell mediated cytotoxicity",
  "effect": "decrease",
  "context": {
    "anatomical_site": "colon tumor",
    "spatial_region": "tumor-immune interface",
    "disease_subtype": "MSI-high"
  },
  "evidence_weight": "high",
  "references": ["PMID:38762484"]
}

What a PhysiCell-style importer would need

• Cell definitions derived from dismech cell type descriptors
• Secreted substrates or state variables derived from cytokines, metabolites, or signaling processes
• Update rules mapping mechanism objects to proliferation, death, motility, exhaustion, recruitment, or secretion parameter changes
• Therapy actions that toggle or attenuate specific mechanism nodes
• Spatial constraints that distinguish core, invasive margin, vasculature, and metastatic niche compartments

Key point

dismech does not need to become the simulator. It needs to become the mechanistic interchange layer between curation and simulation.

Key References

• PhysiCell: An open source physics-based cell simulator for 3-D multicellular systems
• PhysiCell project website
• PhysiGym: bridging the gap between the Gymnasium reinforcement learning application interface and the PhysiCell agent-based model software
• GigaTIME: multimodal AI generates virtual population for tumor microenvironment modeling
• GigaTIME coverage in Microsoft Research 2025 year in review
• TumorTwin: A python framework for patient-specific digital twins in oncology
• Computational Resources for Cancer Research portal
• M4RL: multiscale mathematical model-informed reinforcement learning optimizes combination treatment scheduling in glioblastoma evolution
• SMMART serial-biopsy ecosystem analysis in advanced breast cancer
• SMMART clinical trials research at OHSU

Candidate Diseases for Expansion Beyond CRC

Deliverables This Project Should Produce

1. A TME gap-analysis rubric for cancer entries in dismech
2. A reusable tumor_ecosystem mechanism module
3. A simulator-neutral interaction rule export format
4. A CRC pilot that shows the curation-to-simulation workflow end to end
5. A short list of next cancers where TME enrichment has the highest payoff

Notes

2026-04-12

Project created to evaluate dismech as a mechanistic layer for tumor ecosystem modeling.

Key observations:

1. CRC is the right pilot because the KB already contains a spectrum from baseline adenocarcinoma to MSI-high immune-responsive disease to metastatic seed-and-soil biology.
2. The existing immune_checkpoint_blockade module is already a genuine TME module, just a narrow one centered on adaptive immune resistance.
3. The biggest missing piece is not literature volume but representation: simulators need explicit actors, signals, states, and spatial context.
4. GigaTIME-like systems can help initialize or infer TME state from routine pathology, while PhysiCell, M4RL, and digital twin frameworks can use dismech-derived rules for forward simulation and intervention search.
5. The project should treat raw claim triples as an intermediate product, not the final export. Executable simulation rules need richer structure.
6. First high-confidence attachable PhysiCell anchors for dismech are already available in the official grammar_samples release: epi_caf_invasion and pdac_therapy for PDAC, plus tumor_immune_base and tumor_immune_extended as reusable tumor-immune exemplars.