Prediction of protein loop conformations using multiscale modeling methods with physical energy scoring functions

Area of Science:

• Computational Biology and Biochemistry
• Structural Bioinformatics
• Protein Structure Prediction

Background:

• Accurate prediction of protein loop conformations is crucial for understanding protein function and dynamics.
• Traditional methods often struggle with the conformational flexibility and complexity of protein loops.

Purpose of the Study:

• To develop and evaluate an ab initio computational strategy for predicting protein loop structures.
• To compare the efficiency and accuracy of different physical energy scoring functions and solvent models.

Main Methods:

• A multiscale approach combining lattice-based low-resolution sampling with all-atom refinement simulations.
• Implementation of the replica exchange method for enhanced conformational sampling.
• Application of CHARMM19 and CHARMM22 energy functions with generalized Born (GB) solvent models for scoring.

Main Results:

• Lattice simulations sampled loop conformations with an average global RMSD of 2.21 Å.
• Refinement using CHARMM19/GB improved sampling RMSD to 1.57 Å, outperforming CHARMM22/GB in efficiency.
• Extending all-atom refinement temperatures significantly reduced the RMSD of detected loops.

Conclusions:

• The multiscale computational strategy effectively samples and predicts protein loop conformations.
• CHARMM19/GB offers a more efficient approach for sampling low-RMSD loop basins compared to CHARMM22/GB.
• Optimized simulation parameters, including temperature ranges, enhance the accuracy of loop structure prediction.