Skip to content

anthonyGosme/ontodynamiqueTheory

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

37 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Ontodynamique — Formal System and Empirical Validation

Lean 4 sorry Domains License: CC BY 4.0 Docker Hub

Ontodynamique is a formal ontological framework built from two independent axioms — Axiom I (being = doing: every entity is an act of self-maintenance whose cost is drawn from the very structure it maintains) and Axiom V (exteriority admits degrees) — from which it derives finitude, irreversibility, operational closure, constitutive normativity, and a compositional gradient that classifies any finite system as closure (endogenous cost-bearing), normative carriage (externalized cost), or aggregate (no cycle).

The central empirical prediction is R-XVII: perturbations targeting the structure of a system (its self-maintenance machinery) produce systematically larger displacement than perturbations targeting its input (its metabolic flow), at matched intensity. This asymmetry is indexed on the topological target of the perturbation, not on its amplitude.

📖 Summary Ontodynamique
📖 Manuscript: ontodynamique.com
🧪 Interactive notebook: Google Colab
📄 Empirical paper: arXiv:2512.09352
🔬 DPDR pre-registration: OSF: https://osf.io/unj7f


Cross-Domain Results

The R-XVII signature converges across three causally disjoint biological domains:

Domain Dataset n Cohen's d Ratio S/I p
Gut microbiome MDSINE2 (Gibson et al. 2025) 9 subjects 1.16 1.61× 0.0006
Coral reefs GCBD (van Woesik & Kratochwill 2022) 34,393 obs 0.39 1.80× 1.96 × 10⁻⁴⁸
Cancer pharmacology GDSC (Iorio et al. 2016) 216,764 obs 0.52 1.85× < 10⁻³⁰⁰

Cross-domain convergence: ratio S/I ≈ 1.8×, CV = 7.2%.
Specificity: under intensity-based classification, ratios diverge (CV = 33%). Over 100,000 random binary partitions, none achieves a mean ratio ≥ 1.3 (p < 10⁻⁵).

Validation splits:

  • Coral reefs (CR-02A): temporal split at 2010 — d_TEST = 0.400, ratio S/I = 2.18×, d_TEST within bootstrap CI of TRAIN.
  • Cancer (CR-02B): 70/30 cell-line split, 10 seeds — median ratio = 1.846×, CV = 1.3%, 10/10 significant.

Lean 4 Formalization

The deductive core is mechanized in Lean 4 across 14 files:

  • 504 theorems, 0 sorry, no domain axiom beyond Lean's standard logical axioms (propext, Quot.sound)
  • 2 axioms (I = α + β, V) + 1 corollary (IV)
  • I-γ ("no act without mode"), II, III, VII derived as theorems
  • 10 separating models proving inter-axiom independence (I ⊥ V)
  • 6 separating models proving mutual independence of I-β components
  • R-XVIII asymmetry derived via template_saving

Key files:

File Content
Lean/Autodynamique.lean Structural trunk (101 theorems)
Lean/gradient.lean R-XVII/R-XVIII compositional gradient
Lean/Conscience.lean Subjectivity chain, valence, paliers
Lean/DPDRDerived.lean DPDR predictions (20 theorems)
Lean/InterAxiomIndependence.lean Separating models
Lean/SeparatingModels.lean I-β component independence
Lean/DomainRestriction.lean I-β weakening map (94 theorems classified)
Lean/Bridjes.lean Bridge hypotheses (microbiome, software debt)

Repository Structure

ontodynamiqueTheory/
├── Lean/                      # Lean 4 formalization (14 files)
├── ScriptMDSINE2/             # Microbiome analysis (MDSINE2)
│   ├── 01_phase1_raw_metrics.py
│   ├── 02_phase2_corrected.py
│   ├── 03_phase3_interaction_matrix.py
│   └── 04_robustness_metrics.py
├── ScriptCorail/              # Coral reef analysis (GCBD)
│   ├── corail.py              # Core R-XVII asymmetry test
│   ├── robustness_reef.py     # 4 response transformations
│   └── reef_temporal_split.py # CR-02A temporal validation
├── ScriptGDSC/                # Cancer pharmacology (GDSC)
│   ├── GDSC1.py               # Full analysis with sensitivity sweep
│   ├── GDSC2.py               # Pathway-only classification (v2)
│   └── gdsc_cellline_split.py # CR-02B cell-line validation
├── ScriptCrossDomain/         # Specificity & cross-domain tests
│   ├── SpecifityCheck.py      # 100k permutations, reversibility, target count
│   └── Sentsivity.py          # Sensitivity analysis
├── output/                    # Generated figures and JSON results
├── Ontoaudit4.lean            # Standalone audit file
├── requirements.txt
├── run_all.sh
└── README.md

Quick Start

Option A: Google Colab (recommended)

Open the Colab notebook — it handles all dependencies, data download, Lean installation, and runs every script with full output.

Option B: Docker (recommended for reproducibility)

Docker Hub

Pre-built image with all dependencies (Python 3.10, Lean 4, MDSINE2) — no installation, no restart.

# Run all tests
docker pull anthonygosme/ontodynamique:latest
docker run --rm -v $(pwd)/output:/app/output anthonygosme/ontodynamique

# Run a specific section
docker run --rm -v $(pwd)/output:/app/output anthonygosme/ontodynamique --section gdsc

# Interactive shell
docker run --rm -it -v $(pwd)/output:/app/output anthonygosme/ontodynamique bash

Available sections: lean, mdsine2, gdsc, corail, yeast_hom, yeast_het, crossdomain, meta, artificial

Option C: Local installation

# 1. Clone
git clone https://github.com/anthonyGosme/ontodynamiqueTheory
cd ontodynamiqueTheory

# 2. Python environment
python3 -m venv venv && source venv/bin/activate
pip install -r requirements.txt

# 3. MDSINE2 (microbiome scripts only)
git clone https://github.com/gerberlab/MDSINE2.git
pip install MDSINE2/.
git clone https://github.com/gerberlab/MDSINE2_Paper.git

# 4. Lean 4 (formalization verification)
curl https://raw.githubusercontent.com/leanprover/elan/master/elan-init.sh -sSf | sh -s -- -y --default-toolchain stable
export PATH="$HOME/.elan/bin:$PATH"

# 5. Run everything
bash run_all.sh

Data Sources

Domain Dataset Source
Microbiome MDSINE2 gerberlab/MDSINE2_Paper
Coral reefs GCBD BCO-DMO 773466 (DOI: 10.26008/1912/bco-dmo.773466.2)
Cancer GDSC Sanger dose-response

Robustness

The R-XVII asymmetry has been tested for robustness along multiple axes:

  • Microbiome: 5/5 distance metrics significant (Bray-Curtis, Jensen-Shannon, Aitchison, Hellinger, Canberra; all p < 0.001)
  • Coral reefs: 23/23 DHW thresholds, 9/10 ocean regions, 4/4 response transformations significant; sigmoid threshold stable at DHW ≈ 7.9
  • Cancer: stable from IC30 to IC70; 9/9 pathways significant; dose-matched control preserves ratio (1.81×)
  • Cross-domain specificity: no random partition (n = 100,000) achieves mean ratio ≥ 1.3; reversibility partition dominated 2.8× by R-XVII; target-count partition not operable cross-domain

References

  • Gibson, T.E. et al. (2025). Ecosystem-scale dynamics from microbiome data with MDSINE2. Nature Microbiology.
  • van Woesik, R. & Kratochwill, C. (2022). Global coral bleaching and environmental data. BCO-DMO. DOI: 10.26008/1912/bco-dmo.773466.2
  • Iorio, F. et al. (2016). A landscape of pharmacogenomic interactions in cancer. Cell, 166(3), 740–754.
  • Gosme, A. (2025). Causal symmetrization as empirical signature of operational autonomy. arXiv:2512.09352.
  • Gosme, A. (2026). DPDR protocol — pre-registered. OSF: https://osf.io/unj7f.

Author

Anthony Gosme — Independent researcher
ontodynamique.com

License

This work is licensed under CC BY 4.0.

About

Ontodynamics — Formal axiomatic system (Lean 4, 621 theorems, 0 sorry) with cross-domain empirical programme. Two axioms → operational closure, compositional gradient, subjectivity, normativity.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages