OPUS-Fold3

OPUS-Fold3 performs gradient-based all-atom protein folding and docking to generate and refine three-dimensional protein structures under constraints expressed as potential functions of heavy-atom positions.

Key Features:

• All-atom folding and docking: Implements an all-atom framework that models both backbone and side-chain coordinates for folding and docking tasks.
• Gradient-based optimization: Uses gradient-based methods to optimize atomic positions according to defined potentials.
• Heavy-atom potential functions: Accepts potential functions expressed in terms of the positions of heavy atoms to enforce structural constraints.
• Backbone and side-chain modeling: Performs backbone folding comparable to pyRosetta while demonstrating enhanced side-chain modeling capabilities.
• Implementation: Developed in Python and implemented using TensorFlow 2.4.

Scientific Applications:

• Structural biology: Generation and refinement of accurate 3D protein structures for structural analysis under defined positional constraints.
• Protein design and refinement: Refinement and design of complex protein structures with explicit side-chain modeling.
• Computational drug discovery: Modeling protein dynamics and interactions to support structure-based drug discovery efforts.

Methodology:

Gradient-based optimization of all-atom coordinates using potential functions expressed in heavy-atom positions; explicit modeling of backbone folding and side-chain conformations; implemented in Python with TensorFlow 2.4 and benchmarked against pyRosetta for backbone and side-chain performance.