First-principles protein structure from Horizon-centric Quantum Information and Vacuum (HQIV). No ML, no PDB statistics, no empirical force fields. Formal definitions and proofs live in hqiv-lean.
License: MIT with an explicit government-use restriction — see LICENSE and GOVERNMENT_USE.md. PyPI metadata includes the same notice.
https://disregardfiat.github.io/protein_folder/ — theory (with Lean links), installation, API (mkdocstrings), benchmarks, contributing, live demo.
Build locally: pip install -e ".[dev]" then mkdocs serve.
pip install "protein-folder[full]"Extras: tpu (JAX), grading (pandas), demo (Streamlit + Flask stack). Editable clone:
git clone https://github.com/disregardfiat/protein_folder.git && cd protein_folder
python -m venv .venv && source .venv/bin/activate
pip install -e ".[dev]"Docker: docker build -t protein-folder:cpu . — see Dockerfile and Dockerfile.tpu.
from horizon_physics.proteins import minimize_full_chain, full_chain_to_pdb
result = minimize_full_chain("MKFLNDR", include_sidechains=True)
print(full_chain_to_pdb(result))Package layout: src/horizon_physics/ (import name unchanged). Tests: tests/. CASP-style server: casp_server.py at repo root.
pip install "protein-folder[demo]"
protein-folder-streamlitHQIV framework (academic request in LICENSE): Zenodo.
| Path | Role |
|---|---|
src/horizon_physics/proteins/ |
Folding pipeline, energy, tunnel, CASP hooks |
tests/ |
Pytest + Hypothesis |
docs/ |
MkDocs site sources |
scripts/ |
CASP grading / benchmark drivers |
Details: documentation site and src/horizon_physics/proteins/README.md.
