A branch and bound algorithm for protein structure refinement from sparse NMR data sets

Area of Science:

• Structural biology
• Computational chemistry
• Biophysics

Background:

• Predicting protein tertiary structures is crucial for understanding biological function.
• Nuclear magnetic resonance (NMR) spectroscopy provides distance constraints for structure determination.
• Existing computational methods like distance geometry/molecular dynamics (DGMD) have limitations in accuracy and efficiency.

Purpose of the Study:

• To develop and assess new computational methods for predicting protein tertiary structures to low resolution.
• To improve the accuracy and efficiency of protein structure prediction using NMR data.
• To compare the performance of the new methods against established protocols like X-PLOR.

Main Methods:

• Utilizing secondary structure information and limited long-range NMR distance constraints.
• Employing a modified alphaBB (branch and bound) global optimization algorithm.
• Minimizing an objective function integrating NMR restraints with a protein folding potential (hydrophobicity, excluded volume, van der Waals interactions).

Main Results:

• The new methods achieved substantial improvements in root-mean-square deviation (RMSD) from native structures compared to DGMD.
• DGMD calculations often produced qualitatively erroneous and systematically non-compact structures.
• The developed approach uniformly generated high-quality, low-resolution structures for proteins up to 183 residues with modest computational effort.

Conclusions:

• The new computational/NMR protocol offers a significant advancement in protein structure prediction.
• This methodology provides higher accuracy and reliability than standard DGMD approaches.
• The results encourage the development of protocols for accelerating structure determination in larger, potentially underconstrained proteins.