Research On Protein Structure Prediction And Structure Alignment

One of most important tasks in the post-genome era is inferring function of proteins. Protein function can help us to understand the elusive life system. However, in many cases, illustrating why the protein has such function is more important than the function itself. To deal with this problem, the protein structure should be known. Restricted by the experimental techniques, though many proteins have been sequenced, most of them have not been assigned functional annotations or structures. Bioinformatics, which has been developing rapidly, provides an efficient way to resolve this problem. In the view of this, it's necessary to study the issues related to protein structure, function prediction and analysis using computational methods by characters finding.In this research, a protein secondary structure prediction tool was developed. The position-specific residue preferences around the protein secondary structures' ends were analyzed, and the results showed that there are residues distribution specificity around these sites. Based on this new feature and other features, E-SSpred, a protein secondary structure prediction tool which predicts the secondary structure fragments as a whole, was proposed. E-SSpred was evaluated on standard test datasets and compared with other tools, and the results indicated that E-SSpred can have better performance.By using a novel energy function, a fold recognition method was proposed. The diversity of residue-residue pair-wise interaction in different hydrophobic environment was found. Based on this, a new energy function was proposed and used in fold recognition. The new energy function was tested and compared with common energy function, and the results imply that considering the hydrophobic environment can improve the accuracy of fold recognition.A structure comparison method based on secondary structure elements was proposed. Aimed to resolve the deficiency of recent methods in matching the related residues in query protein and target protein, we improved the secondary structure similarity scoring function and the residue-residue alignment algorithm, and further developed a structure alignment tool, 3D-Sali. The test results indicated that 3D-Sali has good performance on both detecting homology proteins and finding corresponding residues. In the last part, the key factors to decide the amino acid substitution in effecting protein function were analyzed, and were used to improve the prediction. The function sites and their related sites in proteins are found to be more sensitive to impact on the protein function when the substitutions happen on these sites. However, recent widely used features, such as evolution information, cannot show this characteristic. In order to solve this problem, the function annotation database and correlation mutation analysis were used to find the function sites and their related sites, and a method using a novel feature considering these sites was proposed to predict the effect of amino acid substitution to protein function. The test results indicated that this method could improve prediction accuracy effectively.