Reproduce “Agentic End-to-End De Novo Protein Design for Tailored Dynamics Using a Language Diffusion Model” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Bo Ni, Markus J. Buehler
The VibeGen framework, developed by researchers at MIT and Carnegie Mellon, enables end-to-end de novo protein design by explicitly conditioning generation on desired low-frequency normal mode vibrations. The agentic system achieves a median Pearson correlation of 0.72 for smoothed vibration profiles and generates diverse, novel protein sequences with tailored dynamic properties.
Key insights
- Protein language diffusion models can effectively learn and bridge the complex, degenerate relationships between protein sequence, 3D structure, and low-frequency normal mode dynamics.
- An agentic dual-model approach, combining a generative agent (PD) with an evaluative agent (PP), significantly improves the fidelity and efficiency of dynamics-informed protein design.
- The framework implicitly learns how secondary structure elements and local backbone geometries influence vibrational amplitudes, allowing for the design of proteins with tailored flexible or rigid regions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.