OPUS-Rota5: A highly accurate protein side-chain modeling method with 3D-Unet and RotaFormer.

Area of Science:

• Computational Biology
• Structural Biology
• Biochemistry

Background:

• Accurate protein side-chain modeling is essential for predicting protein structure and function.
• Molecular docking relies heavily on precise side-chain interactions for ligand binding.
• Existing methods may struggle with complex side-chain conformations and environmental influences.

Purpose of the Study:

• To introduce OPUS-Rota5, a novel two-stage approach for accurate protein side-chain modeling.
• To evaluate OPUS-Rota5's performance against leading methods on diverse test sets.
• To demonstrate the utility of OPUS-Rota5 in refining protein structures for enhanced molecular docking.

Main Methods:

• Utilized a modified 3D-Unet to extract local residue environmental features, including ligand proximity.
• Employed the RotaFormer module for effective aggregation of diverse feature types.
• Validated the approach on benchmark datasets and recent targets from CAMEO and CASP15.

Main Results:

• OPUS-Rota5 demonstrated significant performance improvements over existing state-of-the-art side-chain modeling techniques.
• Refinement of 25 G protein-coupled receptor targets using OPUS-Rota5 led to a higher success rate in subsequent natural ligand docking.
• The method proved effective for targets with accurate predicted backbones but less accurate side chains.

Conclusions:

• OPUS-Rota5 offers a robust and effective solution for protein side-chain modeling.
• The tool shows particular promise for molecular docking applications, especially with high-homology targets.
• This advancement facilitates more accurate protein design and drug discovery pipelines.