https://github.com/elkins/resonance-flow

JAX-native differentiable protein folding framework integrating experimental NMR constraints (RDCs) and biophysical "self-correction."
https://github.com/elkins/resonance-flow

bioinformatics biophysics differentiable-programming jax machine-learning nmr-spectroscopy protein-folding structural-biology

Last synced: about 2 months ago
JSON representation

JAX-native differentiable protein folding framework integrating experimental NMR constraints (RDCs) and biophysical "self-correction."

• Host: GitHub
• URL: https://github.com/elkins/resonance-flow
• Owner: elkins
• License: mit
• Created: 2026-05-26T22:45:49.000Z (about 2 months ago)
• Default Branch: main
• Last Pushed: 2026-05-26T23:37:49.000Z (about 2 months ago)
• Last Synced: 2026-05-27T00:20:28.890Z (about 2 months ago)
• Topics: bioinformatics, biophysics, differentiable-programming, jax, machine-learning, nmr-spectroscopy, protein-folding, structural-biology
• Language: Python
• Homepage: https://elkins.github.io/resonance-flow/
• Size: 639 KB
• Stars: 1
• Watchers: 0
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# 🧬 ResonanceFlow: Differentiable Protein Structure Prediction

[![PyPI version](https://img.shields.io/pypi/v/resonance-flow.svg)](https://pypi.org/project/resonance-flow/)

[![Supported Python versions](https://img.shields.io/pypi/pyversions/resonance-flow.svg)](https://pypi.org/project/resonance-flow/)

[![Tests](https://github.com/elkins/resonance-flow/actions/workflows/test.yml/badge.svg)](https://github.com/elkins/resonance-flow/actions/workflows/test.yml)

[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)

[![JAX](https://img.shields.io/badge/Accelerated_by-JAX-blue.svg)](https://github.com/google/jax)

[![Code style: black](https://img.shields.io/badge/code%20style-black-000000.svg)](https://github.com/psf/black)

ResonanceFlow is an end-to-end differentiable framework for protein structure prediction that incorporates NMR experimental constraints directly into the folding process.

---

### 🧪 For Structural Biologists

*   **Experimental Self-Correction:** Instead of just predicting a structure, ResonanceFlow uses NMR observables (RDCs, NOEs, Chemical Shifts) to "steer" the model toward the experimental reality.

*   **Physics-Grounded:** Built on the same biophysical kernels used in standard NMR suites, but optimized for the AI era.

### 🤖 For Machine Learning Geeks

*   **Differentiable Physics Loss:** The entire NMR back-calculation is implemented as a differentiable JAX operator, allowing gradients to flow from experimental residuals back to Transformer weights.

*   **Structural Refinement:** Uses a coordination-space predictor that can be fine-tuned on a single protein using only its NMR spectrum as supervision.

---

## 🚀 Key Features

*   **Differentiable NMR Kernels:** Back-calculate RDCs, NOEs, and Chemical Shifts with full gradient support.

*   **Transformer-Based Folding:** Predicts 3D coordinates directly from amino acid sequences.

*   **Self-Correction Loop:** Minimizes the residual between back-calculated and experimental spectra during inference.

## 📦 Installation

```bash

pip install resonance-flow

```

## 📜 License

Distributed under the MIT License. See `LICENSE` for more information.