OPUS-X

OPUS-X predicts protein three-dimensional structures and associated structural features to guide protein folding and structural analysis.

Key Features:

• OPUS-TASS2: Predicts protein torsion angles, secondary structure, and solvent accessibility by incorporating global structural information.
• OPUS-Contact: Computes inter-residue distance and orientation metrics and integrates multiple raw co-evolutionary features with 1D structural predictions to generate refined contact maps.
• OPUS-Fold2: Implements a gradient-based protein folding framework using differentiable energy terms and is implemented in Python with TensorFlow 2.4.
• Performance comparisons: Demonstrates improved contact prediction relative to trRosetta and achieves folding performance comparable to Rosetta when provided identical inputs.
• Modular architecture: Comprises three interrelated modules (OPUS-TASS2, OPUS-Contact, OPUS-Fold2) enabling independent updates of components.

Scientific Applications:

• Structural biology: Supports prediction and analysis of protein 3D structure for structural biology research.
• Protein folding studies: Guides protein folding optimization and assessment using predicted angles, contacts, and differentiable energy terms.
• Drug discovery: Provides structural information and contact maps useful for structure-based drug design and target characterization.
• Functional annotation: Assists functional annotation of proteins through predicted secondary structure, solvent accessibility, and inter-residue relationships.

Methodology:

OPUS-TASS2 predicts torsion angles, secondary structure, and solvent accessibility using incorporated global structural information; OPUS-Contact computes distance and orientation metrics between residue pairs and integrates multiple raw co-evolutionary features with 1D structural predictions to produce refined contact maps; OPUS-Fold2 performs gradient-based folding optimization using differentiable energy terms and is implemented in Python with TensorFlow 2.4.