Chemical shift prediction for protein structure calculation and quality assessment using an optimally parameterized...

Area of Science:

• Structural Biology
• Biophysics
• Computational Chemistry

Background:

• Nuclear Magnetic Resonance (NMR) chemical shifts are sensitive indicators of protein structure.
• Accurate prediction of chemical shifts from protein structures is crucial for structural elucidation.
• Existing formulations for structure-chemical shift relationships have limitations.

Purpose of the Study:

• To present a novel, highly accurate, precise, and robust formulation for predicting NMR chemical shifts from protein structures.
• To improve upon the state-of-the-art in structure-based chemical shift prediction.
• To provide a computationally efficient tool for structural analysis.

Main Methods:

• Developed shAIC (shift prediction guided by Akaike's Information Criterion), a new formulation based on mathematical principles and information theory.
• Represented the structure-chemical shift relationship using a parsimonious sum of smooth analytical potentials.
• Incorporated short-, medium-, and long-range structural parameters in a nuclei-specific manner.
• Optimized the model to capture chemical shift perturbations from distant nuclei.

Main Results:

• shAIC demonstrates superior performance compared to existing analytical formulations for NMR chemical shift prediction.
• The method achieves better results than state-of-the-art approaches, including machine learning methods, for NMR-derived and novel protein structures.
• shAIC offers a computationally lightweight implementation, scalable to large molecules.

Conclusions:

• shAIC provides a significant advancement in predicting NMR chemical shifts from protein structures.
• The formulation's accuracy and robustness make it suitable for various protein structures, including those with novel folds.
• Its computational efficiency positions shAIC as a valuable tool, potentially for use as a force field in molecular simulations.