[Paper Review] Agentic End-to-End De Novo Protein Design for Tailored Dynamics Using a Language Diffusion Model

Bo Ni, Markus J. Buehler|ArXiv.org|Feb 14, 2025

Evolution and Genetic Dynamics3 citations

TL;DR

Introduces VibeGen, an end-to-end de novo protein design framework conditioned on normal mode vibrations, using a dual-model architecture to generate sequences and predict dynamics, validated by full-atom simulations.

ABSTRACT

Proteins are dynamic molecular machines whose biological functions, spanning enzymatic catalysis, signal transduction, and structural adaptation, are intrinsically linked to their motions. Designing proteins with targeted dynamic properties, however, remains a challenge due to the complex, degenerate relationships between sequence, structure, and molecular motion. Here, we introduce VibeGen, a generative AI framework that enables end-to-end de novo protein design conditioned on normal mode vibrations. VibeGen employs an agentic dual-model architecture, comprising a protein designer that generates sequence candidates based on specified vibrational modes and a protein predictor that evaluates their dynamic accuracy. This approach synergizes diversity, accuracy, and novelty during the design process. Via full-atom molecular simulations as direct validation, we demonstrate that the designed proteins accurately reproduce the prescribed normal mode amplitudes across the backbone while adopting various stable, functionally relevant structures. Notably, generated sequences are de novo, exhibiting no significant similarity to natural proteins, thereby expanding the accessible protein space beyond evolutionary constraints. Our work integrates protein dynamics into generative protein design, and establishes a direct, bidirectional link between sequence and vibrational behavior, unlocking new pathways for engineering biomolecules with tailored dynamical and functional properties. This framework holds broad implications for the rational design of flexible enzymes, dynamic scaffolds, and biomaterials, paving the way toward dynamics-informed AI-driven protein engineering.

Motivation & Objective

• Motivate design of proteins with tailored dynamical properties linked to function.
• Develop an end-to-end generative framework conditioned on vibrational modes.
• Bridge sequence design with dynamic accuracy via bidirectional assessment.

Proposed method

• Propose VibeGen, a generative AI framework conditioned on normal mode vibrations.
• Use an agentic dual-model system: a protein designer to generate sequences and a protein predictor to evaluate dynamics.
• Leverage full-atom molecular simulations for direct validation of prescribed normal mode amplitudes.
• Demonstrate that de novo sequences reproduce target backbone dynamics and adopt stable, functional structures.
• Show that generated sequences are novel and lack significant similarity to natural proteins.

Experimental results

Research questions

• RQ1 Can end-to-end de novo protein design be conditioned on prescribed vibrational modes?
• RQ2 Do jointly optimized designer and predictor models achieve accurate dynamic realization in generated sequences?
• RQ3 Are the generated, vibration-conditioned proteins novel relative to natural proteins?
• RQ4 Can full-atom simulations validate that designed proteins reproduce intended normal mode amplitudes?
• RQ5 What is the impact of integrating protein dynamics into generative design for flexible enzymes and dynamic biomaterials?

Key findings

• The framework enables end-to-end design conditioned on targeted normal mode amplitudes.
• Designed proteins reproduce prescribed backbone vibrational amplitudes in simulations.
• Generated sequences are de novo with no significant similarity to natural proteins.
• Designed structures are stable and functionally relevant across modes.
• The approach links sequence directly to vibrational behavior, expanding design space beyond evolution.