Molecular Dynamics-Assisted Optimization of Protein NMR Relaxation Analysis

Area of Science:

• Biophysics
• Computational Biology
• Structural Biology

Background:

• NMR relaxation analysis relies on internal autocorrelation functions to model protein residue motion.
• Existing order parameter representations can fit experimental data but yield divergent motion predictions.
• Distinguishing physically realistic models from experimental data alone is challenging.

Purpose of the Study:

• To evaluate different order parameter representations for modeling protein backbone dynamics.
• To assess the capability of molecular dynamics simulations in discriminating these representations.
• To compare the performance of various AMBER and CHARMM force fields in predicting ubiquitin's conformational dynamics.

Main Methods:

• Calculated autocorrelation functions for ubiquitin's backbone H-N bond vectors using six AMBER and CHARMM force fields.
• Compared the optimized time constant-constrained triexponential (TCCT) representation against extended Lipari-Szabo and Larmor frequency-selective representations.
• Utilized experimental 15N relaxation data (R1, R2, NOE) to validate force field predictions.

Main Results:

• The TCCT representation significantly outperformed other models in accurately predicting autocorrelation functions.
• Optimization of TCCT parameters converged to the same values at 600 and 900 MHz 1H.
• Higher magnetic fields offered limited additional benefit for mobile amides without slow exchange effects.
• Experimental 15N relaxation data effectively differentiated force field accuracy for ps-ns backbone dynamics.
• AMBER 14SB demonstrated the most consistent predictions for ubiquitin's mobile C-terminal residues, unlike older force fields that underestimated dynamics.

Conclusions:

• Molecular dynamics simulations combined with NMR relaxation data are crucial for selecting accurate models of protein dynamics.
• The TCCT representation provides a more physically realistic description of protein motion than widely used alternatives.
• AMBER 14SB is recommended for accurate modeling of ubiquitin backbone dynamics in the ps-ns timescale.