The Research On The Electron-density Map Interpretation Based On The Knowledge Of Genetically Conserved Structure Model

An important problem in high-throughput protein crystallography is constructing a protein model from an electron-density map. This is a process of matching a whole protein molecule model into a crystal protein3D image. In poor-quality electron density maps, the interpretation may require a significant amount of a crystallographer’s time.The main research work of this thesis is to propose an ideal method to work well in poor-quality density maps, which takes shorter program execute time, higher precision of analytical results. By analyzing the existing methods, we know that the classical ACMI algorithm has a better performance, but repeated calculation will happened inevitably since the unite of search for this method is small template fragments, and it has a large number of times of searching, which leads to higher complexity of calculating time.In this thesis, we propose a new improved method based on ACMI algorithm using the knowledge of genetically conserved structure model. According to the protein amino acid sequence to be tested, we search out the possible corresponding template with the sequence in protein database (PDB) base on the knowledge of protein structure maintenance and use the method of sequence alignment, and then locate the corresponding template’s spatial structure into the optimal matching position of electron-density map by using the method of spherical harmonics and fast template matching. Comparing to search and location using5residues as a unit, this kind of strategy of locate larger template saves a lot of time. Next, in the process of template connecting, improving and refining, we use ACMI algorithm again to locate some small residues to electron-density map through an effective approximate-inference algorithm—Belief Propagation.We have test the algorithm of this research through a series of poor-quality protein electron-density map (3.5angstrom resolution), and compared our results with ACMI. Compared with ACMI, the algorithm proposed in this research provides more precise backbone trace, which not only improves precision but also shortens the executing time in the process of constructing protein model.