OPUS-PSP: an orientation-dependent statistical all-atom potential derived from side-chain packing.

Area of Science:

• Computational Biology
• Structural Biology
• Biophysics

Background:

• Protein structure prediction is a fundamental challenge in molecular biology.
• Accurate modeling of protein structures requires precise representation of atomic interactions.
• Knowledge-based potentials are widely used in protein structure prediction.

Purpose of the Study:

• To develop a novel all-atom statistical potential for protein structure modeling.
• To capture orientation-dependent interactions crucial for native protein structures.
• To improve the accuracy of protein structure prediction and refinement.

Main Methods:

• Developed OPUS-PSP, an orientation-dependent statistical potential based on side-chain packing.
• Utilized a basis set of 19 rigid-body blocks representing amino acid residues.
• Generated the potential from orientation-specific packing statistics in a non-redundant structural database.

Main Results:

• OPUS-PSP significantly outperforms existing knowledge-based potentials on decoy sets.
• Demonstrated superior ability in recognizing native protein structures.
• Achieved consistent high Z-scores across various decoy sets, indicating robustness.

Conclusions:

• Side-chain packing is critical for forming native protein structures, as highlighted by OPUS-PSP's success.
• OPUS-PSP's exclusion of explicit solvation terms makes it suitable for modeling membrane proteins.
• The potential is broadly applicable for protein structure modeling, especially for side-chain conformations.