Determination of three-dimensional structure of protein is an important task in understanding their exact function and role in organism. More than 138,000 protein structures determined by X-ray, NMR (Nuclear Magnetic Resonance) are deposited in PDB (P...
Determination of three-dimensional structure of protein is an important task in understanding their exact function and role in organism. More than 138,000 protein structures determined by X-ray, NMR (Nuclear Magnetic Resonance) are deposited in PDB (Protein Data Bank). There are about 11% of NMR structures in the whole structures of PDB and although the ratio increases gradually over the years, it still has a poor quality compared to the X-ray structure. The low quality of the NMR structure can be a hindrance to future drug development and structural research. Therefore, refinement is a necessary step for structural research using NMR structures. In this study, we introduce a method to improve the structural quality of NMR protein structure. Some of the problems of NOE experimental data obtained from NMR experiments were overcome and used for refinement. First, due to the ambiguity of NOE, it cannot be used immediately for structure calculation, and it requires handling procedure. The ambiguity is classified into four types and used for simulation after handling. Second, NMR structures with no NOE registered in the BMRB are difficult to improve. In order to improve the NMR structure regardless of the presence of NOE, a structure-derived distance restraint created from the hydrogen atom distance information obtained from a given structure was used in place of NOE to improve the structure. With these geometrical restraints, the statistical torsion angle potential (STAP) developed from the torsion angle population of high resolution X-ray structures was used. This contributed to the generation of secondary structures and improved Ramachandran accuracy. The refined structure has been measured for structural quality through various criteria: satisfaction of experimental data, geometrical/stereo-chemical protein quality, ensemble conservation degree and secondary structure ratio. After refinement, the refined structures were satisfied with the experimental data by having a much lower violation value than initial structure. The refinement protocol is built with a web server (NMRe), which allows the user to directly input the NMR structure to obtain refined structure, as well as providing various structural analysis results. The NMR structures with better reliability and accuracy than the initial structure can be expected to be more actively used in structural bioinformatics such as structure-based drug development and protein design.