OPUS-Fold3: a gradient-based protein all-atom folding and docking framework on TensorFlow.

Area of Science:

• Computational Biology
• Structural Bioinformatics
• Artificial Intelligence in Biology

Background:

• Accurate protein structure prediction is crucial for molecular design and function analysis.
• Existing protein folding and docking frameworks face limitations in efficiency and accuracy, particularly for complex modeling tasks.

Purpose of the Study:

• To introduce OPUS-Fold3, a novel gradient-based, all-atom framework for protein folding and docking.
• To demonstrate the framework's capability in generating accurate 3D protein structures under specified constraints.
• To provide a user-friendly and adaptable tool for the scientific community.

Main Methods:

• Development of OPUS-Fold3 using Python and TensorFlow 2.4.
• Implementation of a gradient-based, all-atom approach for protein structure modeling.
• Constraint-based structure generation utilizing potential functions dependent on heavy atom positions.

Main Results:

• OPUS-Fold3 demonstrates performance comparable to pyRosetta in backbone folding.
• The framework shows significantly superior performance in side-chain modeling compared to pyRosetta.
• Successful generation of 3D protein structures compliant with user-defined constraints.

Conclusions:

• OPUS-Fold3 offers an efficient and accurate solution for protein folding and docking.
• The framework's flexibility facilitates integration with other deep learning models, fostering interdisciplinary research.
• OPUS-Fold3 is freely available for academic use, promoting advancements in structural biology and AI.