Assessing the quality of solution nuclear magnetic resonance structures by complete cross-validation

Area of Science:

• Biochemistry
• Structural Biology
• Spectroscopy

Background:

• Macromolecular structure determination in solution relies on fitting atomic models to nuclear magnetic resonance (NMR) data, specifically nuclear Overhauser effect (NOE) observations.
• Assessing the quality of these NMR-derived structures is crucial for reliable interpretation.

Purpose of the Study:

• To establish reliable and unbiased criteria for evaluating the quality of solution NMR structures.
• To investigate the effectiveness of complete cross-validation for structure quality assessment.

Main Methods:

• Utilized complete cross-validation by partitioning nuclear Overhauser effect (NOE) data into test sets.
• Evaluated statistical quantities derived from these test sets to assess structural fit.
• Correlated cross-validated measures (e.g., distance bound violations, NMR R values) with structure quality.

Main Results:

• A high correlation was observed between cross-validated measures of fit and the actual quality of solution NMR structures.
• Incomplete data sets led to poorer satisfaction of cross-validated quality metrics.
• The study demonstrated the utility of cross-validation for quality control in NMR structure determination.

Conclusions:

• Complete cross-validation offers a robust method for defining reliable criteria for solution NMR structure quality.
• Optimizing cross-validated measures of fit is expected to yield solution NMR structures with maximum information content.
• This approach enhances the accuracy and reliability of structural biology research using NMR.