Technical Highlight - April 2012
Short description: A rigorous assessment of automated NMR structure determination demonstrates the reliability of the method.
Protein structure determination by NMR spectroscopy requires complete chemical shift assignments and manual analysis of thousands of atomic interactions known as nuclear Overhauser effects (NOEs). Recently developed algorithms use these data to calculate structures in automated, iterative NOE assignment protocols. Additional methods have emerged that rely on the intrinsic structural information encoded in the chemical shifts supplemented with NOE data.
Rosato and colleagues (PSI NESG) present the results of CASD-NMR, a blind assessment of how the available automated methods compare with the traditional manual approach. The authors stringently evaluated the results for ten structure determinations of monomeric proteins ranging in size from 60 to 150 amino acids. They assessed three groups of automated methods using either NOE or chemical shift data, or a combination of both. The accuracy and convergence of the automatically calculated structures was compared to manually solved reference structures using the root-mean-square deviation (RMSD) of structured backbone regions.
Setting a threshold for accuracy at an RMSD of 2 Å from the reference structures, the authors found that three of the four NOE-based methods consistently generated accurate structures for at least 90% of the targets. Methods that primarily use chemical shift data supplemented with or filtered by NOE restraints yielded acceptable structures in 70% of cases, while computationally expensive methods that use only chemical shift data performed poorly. Using crystal structures of homologs for two of the targets, the authors were able to construct reliable models with acceptable RMSD deviations to both automated and reference structures. In one notable case, both the automated and reference structures were closer to the DNA-bound crystal structure, indicating that the free protein in solution also adopts the bound conformation.
Further analysis revealed that correct geometric and stereochemical parameters are necessary but not sufficient to guarantee accurate structures. The authors propose an alternative quality measure that quantifies the agreement with the experimental data. Overall, the study has established the reliability of NOE-based methods in unsupervised automated solution structure determination of proteins of up to 150 amino acids in size.
A. Rosato et al. Blind testing of routine, fully automated determination of protein structures from NMR data.
Structure 20, 227-236 (2012). doi:10.1016/j.str.2012.01.002