Accurate Structure Prediction for Protein Loops Based on Molecular Dynamics Simulations with RSFF2C

Area of Science:

• Computational biology
• Structural biology
• Biophysics

Background:

• Protein loops connect secondary structures (α-helices and β-strands) and are crucial for biological functions.
• The conformational flexibility of protein loops makes their accurate three-dimensional (3D) structure determination a significant challenge, both experimentally and computationally.

Purpose of the Study:

• To systematically evaluate computational methods for predicting the 3D structures of long, solvent-exposed protein loops.
• To identify the most effective molecular dynamics (MD) simulation force fields and enhanced sampling techniques for accurate protein loop structure prediction.

Main Methods:

• Evaluated four leading loop modeling algorithms (DaReUS-Loop, Sphinx, Rosetta-NGK, MODELLER) on 15 long loops.
• Performed extensive molecular dynamics (MD) simulations, including temperature replica exchange molecular dynamics (REMD), using three force fields: RSFF2C, CHARMM36m, and AMBER ff19SB.
• Investigated normal MD simulations at elevated temperatures (380 K, 500 K, 620 K) as a lower-cost alternative.

Main Results:

• State-of-the-art loop modeling algorithms showed limited success in accurately predicting loop structures.
• The residue-specific force field RSFF2C, combined with the TIP3P water model, demonstrated superior performance, accurately predicting 12 out of 15 loops (RMSD < 1.5 Å) using REMD simulations.
• High-temperature MD simulations with RSFF2C+TIP3P yielded results comparable to REMD, suggesting potential for reduced computational cost.

Conclusions:

• Molecular dynamics simulations, especially with enhanced sampling techniques like REMD, coupled with the RSFF2C force field, offer a promising approach for accurate protein loop structure prediction.
• The RSFF2C force field represents a significant advancement for modeling the conformational dynamics of flexible protein loops.
• High-temperature MD simulations present a viable, computationally less intensive alternative for loop structure prediction using optimized force fields.