Improved Sampling Strategies for Protein Model Refinement Based on Molecular Dynamics Simulation.

Area of Science:

• Computational Biology
• Structural Biology
• Biophysics

Background:

• Protein structures are crucial for understanding biological functions.
• Experimental methods like X-ray crystallography determine protein structures, but in silico prediction is an alternative.
• Current in silico methods, including template-based modeling and machine learning, face accuracy challenges for proteins with no known structural homologs.

Purpose of the Study:

• To enhance the accuracy of predicted protein structures using physics-based refinement.
• To explore novel simulation schemes for improving protein model quality.
• To develop a superior refinement protocol for in silico protein structure prediction.

Main Methods:

• Utilized molecular dynamics simulations with optimized biasing functions and elevated temperatures.
• Employed alternative initial models to broaden conformational space exploration.
• Developed and benchmarked a new simulation-based refinement protocol.

Main Results:

• The proposed refinement protocol significantly improved protein model accuracy compared to existing methods.
• Optimized biasing functions and increased simulation temperatures enhanced conformational sampling.
• Broader exploration of conformational space through alternative initial models contributed to improved refinement.

Conclusions:

• The novel refinement protocol offers a substantial advancement in in silico protein structure prediction.
• This approach addresses limitations in current molecular dynamics-based refinement techniques.
• The findings pave the way for more accurate prediction of protein structures, particularly for challenging targets.