AlphaDynamics: Compact Phase-Oscillator Neural Propagator for Protein Torsion Dynamics with Backbone 3D Reconstruction

Fuente: Zenodo
Saved in:
Bibliographic Details
Main Author: Gwozdz, Krzysztof
Format: Recurso digital
Published: Zenodo 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866902102224142336
author Gwozdz, Krzysztof
author_facet Gwozdz, Krzysztof
contents AlphaDynamics v0.4.0 — Python package for predicting protein backbone torsion dynamics from sequence using a compact (78K params) phase-oscillator neural propagator. Achieves 2.39x lower mean Ramachandran JSD than Microsoft Timewarp 4AA model with 9,652x fewer parameters. New in v0.4.0: deterministic backbone 3D reconstruction from torsion trajectories using NeRF (Parsons 2005) with Engh-Huber 1991 standard bond geometry. Provides multi-model PDB output for visualization in PyMOL/VMD/ChimeraX, plus per-frame radius of gyration and end-to-end distance diagnostics. Scope note: best-validated for short peptides (4-15 aa); sequences longer than 20 aa are outside the calibrated scope.
format Recurso digital
id zenodo_https___doi_org_10_5281_zenodo_20100149
institution Zenodo
language
publishDate 2026
publisher Zenodo
record_format zenodo
spellingShingle AlphaDynamics: Compact Phase-Oscillator Neural Propagator for Protein Torsion Dynamics with Backbone 3D Reconstruction
Gwozdz, Krzysztof
protein dynamics
molecular dynamics
torsion angles
phase oscillators
neural ODE
Ramachandran
backbone reconstruction
NeRF
mdCATH
Timewarp
ml-for-biology
AlphaDynamics v0.4.0 — Python package for predicting protein backbone torsion dynamics from sequence using a compact (78K params) phase-oscillator neural propagator. Achieves 2.39x lower mean Ramachandran JSD than Microsoft Timewarp 4AA model with 9,652x fewer parameters. New in v0.4.0: deterministic backbone 3D reconstruction from torsion trajectories using NeRF (Parsons 2005) with Engh-Huber 1991 standard bond geometry. Provides multi-model PDB output for visualization in PyMOL/VMD/ChimeraX, plus per-frame radius of gyration and end-to-end distance diagnostics. Scope note: best-validated for short peptides (4-15 aa); sequences longer than 20 aa are outside the calibrated scope.
title AlphaDynamics: Compact Phase-Oscillator Neural Propagator for Protein Torsion Dynamics with Backbone 3D Reconstruction
topic protein dynamics
molecular dynamics
torsion angles
phase oscillators
neural ODE
Ramachandran
backbone reconstruction
NeRF
mdCATH
Timewarp
ml-for-biology
url https://doi.org/10.5281/zenodo.20100149