OPUS-Fold

OPUS-Fold implements torsion-angle sampling and composite energy evaluation to reconstruct and optimize all-atom protein structures for structure prediction.

Key Features:

• Torsion-angle sampling: Uses torsion-angle sampling as the core methodology to explore backbone conformations and enable reconstruction after each sampling step.
• Structure reconstruction: Reconstructs protein structures incrementally following each sampling step to produce models suitable for scoring.
• Composite energy function: Evaluates model quality with a multi-term energy function that combines constraint terms developed by the OPUS-Fold team and terms from the literature.
• OPUS-Rota2 side-chain modeling: Integrates the OPUS-Rota2 fast side-chain modeling method to construct all-atom atomic models during folding.
• Integration of subproblem methods: Supports incorporation of diverse subproblem approaches, including those that require all-atom data.
• Support for nondifferentiable and experimental data: Accepts nondifferentiable information such as partial experimental data for refining structure predictions.
• Computational efficiency: Balances accuracy with computational efficiency to enable incorporation of additional subproblem methods into the folding process.

Scientific Applications:

• Protein structure prediction: Produces and optimizes predicted protein structures using torsion-angle sampling and energy-based scoring.
• All-atom model generation: Generates all-atom models with side chains to support downstream analyses and subproblem methods that require detailed atomic representation.
• Refinement with experimental data: Refines structure predictions by integrating partial experimental data and nondifferentiable constraints alongside computational methods.

Methodology:

Torsion-angle sampling with incremental reconstruction after each sampling step; model scoring by a multi-term composite energy function combining OPUS-Fold and literature-derived terms; OPUS-Rota2 fast side-chain modeling for all-atom model construction; support for integration of nondifferentiable partial experimental data.