New Approach To Protein Structure Calculation From NMR Data

Knowing the structure of proteins is critical for understanding their physical and chemical properties. A major problem in bioinformatics is to efficiently determine the three-dimensional structure of a protein. Protein NMR structure determination is one of the main experimental methods. In protein NMR, the three-dimensional structure is determined by exploiting a set of distance restraints between spatially proximate atoms. Up to now, there is no practical automated protein NMR method without human intervention.Describing protein structure by distance geometry is very important. In this paper, we propose a two-stage method(TSM) to estimate the structure of a protein from highly incomplete NMR distance matrix. We first randomly “guess” a small part of unobservable distances from observable distances by utilizing the triangle inequality, which is crucial for the second stage. Matrix Completion(MC) is used to estimate the protein structure from the obtained distance matrix. An accelerated proximal gradient algorithm(APG) is applied to solve the corresponding optimization problem. As example, four protein structures are calculated by TSM, which presented its efficiency in the recovery of protein structure. Finally the recovery error is also analyzed. Four protein structures are calculated by SPROS. Its result shows the practical of our approach.