Comparison between Generalized-Born and Poisson-Boltzmann methods in physics-based scoring functions for protein...

Area of Science:

• Computational chemistry
• Structural biology
• Biophysics

Background:

• Continuum solvent models like Generalized-Born (GB) and Poisson-Boltzmann (PB) are crucial for efficient solvation effect treatment in protein structure scoring.
• A direct comparison of GB and PB methods on large protein datasets is currently lacking.
• Previous work established a scoring function using a GB solvation model and molecular dynamics simulations.

Purpose of the Study:

• To extend a GB-based scoring function and compare its performance with the MM-PBSA method for treating solvent effects.
• To benchmark the developed scoring function against established protein decoy sets.
• To analyze the impact of large ligands and ions on scoring accuracy.

Main Methods:

• Development of an extended scoring function incorporating the MM-PBSA method for solvent effect calculation.
• Benchmarking the scoring function against seven publicly available protein decoy sets.
• Analysis of scoring accuracy in the presence of large ligands and ions in native structures.

Main Results:

• The MM-PBSA approach demonstrated comparable performance to the previously developed GB-based scoring function.
• The study provides a direct comparison of GB and MM-PBSA on a large scale.
• The influence of ligands and ions on scoring accuracy was investigated.

Conclusions:

• Poisson-Boltzmann (PB) and Generalized-Born (GB) continuum solvent models show comparable efficacy in protein structure scoring when integrated with physics-based energy functions.
• The findings suggest that MM-PBSA offers a viable alternative to GB models for protein structure evaluation.
• Further investigation into the effects of ligands and ions is warranted for refining scoring function accuracy.