T1D Digital Twin & Hypoimmune Organoid Simulator

Language / Язык: English · Русский

Open-source, literature-calibrated engine (t1d_simulator) by Pavel V. Naumov. It maps which implant geometries for encapsulated $\beta$-cells are physically allowed by oxygen transport — and which are not — before wet-lab work.

Main result: a checked Krogh length $L^$. Thick subcutaneous slabs ($\ge 250\ \mu\mathrm{m}$ half-thickness at $\ge 50$ M/ml) and $400\ \mu\mathrm{m}$ capsules at typical density sit above $L^$. A thin omental sheet or small spheres do not. Edmonton-like cell mass is an area problem (~650 cm²), not a thicker pancake. Not a clinical product.

---

🔮 Core Features

• Oxygen Transport PDE Solver: Solves the non-linear reaction-diffusion equations under spherical, cylindrical, and planar geometries to identify diffusion barriers and necrotic core formation ($pO_2 < 0.5\ \mathrm{mmHg}$).
• Physics-Informed Neural Network (PINN): A mesh-free neural solver surrogate that computes steady-state oxygen concentration profiles in under 10 milliseconds, enabling real-time parameter sweeps.
• IBMIR Coagulation ODE Cascade: A compartmental model of the 0–48h Instant Blood-Mediated Inflammatory Reaction (Tissue Factor → Thrombin → Fibrin Clot) which dynamically reduces oxygen permeability at the boundary.
• GNN Anti-Fibrotic Screening: A Graph Neural Network that maps chemically modified alginate monomers (SMILES) to foreign body reaction (FBR) fibrotic capsule thickness ($L_{\mathrm{fib}}$) predictions.
• Clinical Dose Calculator: Translates virtual patient mass and daily insulin requirements (TDI) into target islet equivalents (IEQ), organoid counts, and omental scaffold volume.
• Bergman OGTT Glycemic Dynamics: Simulates glucose-stimulated insulin secretion (GSIS) and glucose clearance curves during oral glucose tolerance tests (OGTT).
• CGM 30-Day Monitoring: Simulates long-term continuous glucose metrics, returning GMI (HbA1c), Time-in-Range (TIR), Time-Below-Range (TBR), and Time-Above-Range (TAR).
• CAD/STL Scaffold Exporter: Procedurally generates and exports 3D porous TPMS scaffolds (STL format) matching patient-specific omental pouch area requirements.

---

📂 Repository Structure

.
├── .github/workflows/          # CI/CD pipelines
├── data/
│   ├── literature_params.yaml  # Curated parameters with literature citations
│   └── benchmarks.py           # Verification script for validation datasets
├── docs/
│   ├── figures/                # Manuscripts schematics and graphical abstract
│   ├── manuscript_biorxiv_en.md  # Preprint v1.3 (English)
│   ├── manuscript_biorxiv_ru.md  # Preprint v1.3 (Russian)
│   └── endocrinologist_brief_ru.md
│   ├── site_comparison_matrix.md# Comparative matrix of transplant locations
│   └── *.md                    # Clinical passports, MoU templates, protocols
├── reports/
│   ├── benchmarks/             # Predicted vs reported validation
│   └── sweeps/                 # screen_design --sweep CSV/JSON/heatmaps
├── archive/                    # Historical notes (not the public product)
├── scripts/                    # Helper scripts for screening and email templates
├── t1d_simulator/              # Python simulator source package
│   ├── app.py                  # Main Streamlit dashboard (monolith)
│   ├── ibmir_module.py         # IBMIR kinetics solver
│   ├── verify_model.py         # Integration tests (44)
│   └── parameters.yaml         # Default runtime parameters
└── pyproject.toml              # Build configurations and dependency lists

---

🚀 Quickstart

1. Installation

The package runs on Python 3.9–3.12. Clone the repository and install dependencies:

git clone https://github.com/Bezooom/t1d-simulator.git
cd t1d-simulator
pip install -r t1d_simulator/requirements.txt

2. Run Interactive Dashboard

Launch the Streamlit graphical user interface:

streamlit run t1d_simulator/app.py

3. Verify Model and Run Tests

Run the 44-test integration test suite on CPU:

export CUDA_VISIBLE_DEVICES=""
python3 t1d_simulator/verify_model.py

4. Execute Construct Screening

Single design:

python3 t1d_simulator/screen_design.py --geometry spherical --radius 200 --density 80 --site omental_pouch

Density × thickness sweep + heatmap (the design-space product):

python3 t1d_simulator/screen_design.py --sweep \
  --geometry planar --site subcutaneous --no-heparin \
  --densities 20,50,80,100 --radii 100,150,200,250 \
  --export-csv reports/sweeps/planar_sq.csv \
  --export-json reports/sweeps/planar_sq.json \
  --heatmap reports/sweeps/heatmap_viability_planar_sq.png \
  --heatmap-po2 reports/sweeps/heatmap_po2_planar_sq.png

5. Verified design bounds (analytical Krogh vs BVP)

CUDA_VISIBLE_DEVICES="" python3 t1d_simulator/design_bounds.py

Report: reports/design_bounds/verified_design_space.md.

Dose vs Krogh (Edmonton mass → surgical area): python3 t1d_simulator/dose_transport.py → reports/dose_transport/dose_vs_krogh.md.

Islet scale + death window + literature sizes: python3 t1d_simulator/scaleup_science.py → reports/scaleup/remaining_biotech_bounds.md.

---

Documentation

	English	Русский
This README	README.md	README.ru.md
Preprint	manuscript_biorxiv_en.md	manuscript_biorxiv_ru.md
One page for clinicians	—	endocrinologist_brief_ru.md
Geometry bound $L^*$	verified_design_space.md	same numbers
Dose → area	dose_vs_krogh.md	same
Islet size + death window	remaining_biotech_bounds.md	same
Who to write (RU-first)	—	outreach_playbook.md
Docs index	docs/README.md	docs/README.md

---

📊 Scientific Validation

The digital twin's transport and cell survival solvers have been benchmarked and validated against independent published datasets:

Benchmark	Reference Paper	Target Metric	Agreement Status
Krogh Limit	Secomb et al. (2004)	Viability boundary limit (~200 $\mu\text{m}$)	PASS (1.00 ratio)
Spherical Alginate	Papas et al. (2007)	Viability sweep ($R = 400\ \mu\text{m}$)	PASS (RMSE = 10.42%)
Planar Macrodevice	Papabathini et al. (2023)	Hypoxic core profile ($L = 250\ \mu\text{m}$)	PASS (RMSE = 11.42%)
IBMIR Time-Course	Hackett et al. (2013)	48h acute cell retention	PASS (1.03 ratio)
Angiogenesis	Berney et al. (2016)	Capillary perfusion time	PASS (0.80 ratio)

---

📑 Citation & Academic Use

If you use this simulator or reference the preprint findings in your research, please cite the work as follows:

@article{naumov2026multiphysics,
  title={Multiphysics Failure Modes of \beta-Cell Encapsulation: From the Krogh Diffusion Limit to Hypoimmune Organoid Digital Twins},
  author={Naumov, Pavel V.},
  journal={bioRxiv},
  year={2026},
  doi={To be assigned},
  url={https://github.com/Bezooom/t1d-simulator}
}

---

📄 License

This project is licensed under the MIT License — see the LICENSE file for details. Open for academic and commercial co-development. For research collaborations and custom in silico design sweeps, please contact naumov122@gmail.com.