All-Atom Four-Body Knowledge-Based Statistical Potentials to Distinguish Native Protein Structures from Nonnative...

Area of Science:

• Computational Biology
• Structural Biology
• Biophysics

Background:

• Protein folding is crucial for biological function, and accurate structure prediction remains a challenge.
• Coarse-grained models and empirical energy functions have advanced protein structure understanding.
• Distinguishing native protein structures from nonnative decoys is essential for reliable prediction.

Purpose of the Study:

• To develop and evaluate novel all-atom four-body statistical potentials for protein structure prediction.
• To assess the performance of these potentials against existing physics- and knowledge-based models.
• To demonstrate the potential's utility in calculating binding energies for protein-ligand complexes.

Main Methods:

• Utilized atomic coordinates from a diverse protein chain training set.
• Developed twelve all-atom four-body statistical potentials by varying key parameters.
• Employed Delaunay tessellation to identify interacting atom quadruplets.
• Applied statistical analysis and the inverted Boltzmann principle to generate atomic potentials.
• Evaluated potentials using the Decoys-'R'-Us benchmarking dataset.

Main Results:

• The best developed potential ranked third overall, matching CHARMM19 and exceeding AMBER force field potentials.
• Demonstrated the potential's effectiveness in distinguishing native protein structures from decoys.
• Successfully applied a generalized version to calculate binding energies for HIV-1 protease-inhibitor complexes.

Conclusions:

• The developed atomic four-body statistical potential offers an accurate and efficient method for protein structure prediction and assessment.
• This potential provides a valuable tool for computational biology and drug discovery.
• The findings highlight the significance of four-body interactions in accurately modeling protein structures.