Physical scoring function based on AMBER force field and Poisson-Boltzmann implicit solvent for protein structure...

Area of Science:

• Computational Biology
• Biophysics
• Structural Bioinformatics

Background:

• Protein structure prediction is crucial for understanding biological function.
• Physics-based scoring functions aim to accurately model molecular interactions.
• Existing methods face challenges in discriminating native structures from decoys.

Purpose of the Study:

• To analyze the performance of the AMBER/Poisson-Boltzmann (PB) physics-based scoring function.
• To evaluate its ability to discriminate native protein structures from decoys.
• To identify key components influencing the scoring function's discriminative power.

Main Methods:

• Utilized the AMBER force field for intramolecular interactions and an efficient Poisson-Boltzmann (PB) model for solvation.
• Tested the scoring function on diverse all-atom decoy sets, including crystal and NMR structures.
• Analyzed the contribution of different scoring function components to ranking native and near-native structures.

Main Results:

• The AMBER/PB scoring function successfully discriminated all 62 native crystal structures, accounting for missing details like heteroatoms and disulfide bonds.
• It achieved 80% accuracy for NMR structures.
• The function effectively selected near-native decoys, as indicated by ranking numbers and Calpha root mean square deviations.
• Nonpolar solvation energy and higher protein dielectric constants did not enhance discriminative power.
• 1-4 electrostatics and dihedral-angle terms were identified as areas for improvement, with 1-4 electrostatics being the most problematic.

Conclusions:

• The AMBER/PB scoring function demonstrates robust performance in protein structure prediction.
• Physics-based potentials are becoming competitive with knowledge-based potentials.
• Further refinement of terms like 1-4 electrostatics is necessary for enhanced accuracy.