Mathematical Methods In The Application Of The Protein Sequence Analysis

Membrane proteins play several roles in the cells. They are involved in almost allcellular processes such as communication with the outside world, transport of nutrientand metabolism across membranes, and energy generation processes. Membraneproteins are difficult to study mostly because they are often poorly abundant and thusdifficult to purify in amounts compatible with structural studies. Thus, prediction ofmembrane protein structures by the methods of bioinformatics, mathematics andcomputer simulating is a fundamental work to find and study the structure andfunction for new transmembrane protein.In the thesis, the helix structure of transmembrane protein is introduced. Theconcept of weight is suggested to predict the transmembrane domains oftransmembrane proteins. Firstly, the hydrophobic parameters of20amino acids areweighted according to our method. By the new hydrophobic parameters of each aminoacid, transmembrane domains of transmembrane proteins are predicted. Theprediction accuracy of residue is52.92%and prediction accuracy for transmembranedomains is88.87%in this thesis.It is worthy of note that the interaction of neighbor amino acids plays a major rolein molding protein structure. In this thesis, considering the interaction betweenneighbor amino acids, the3D graphical representation for a protein is proposed. Then,a novel descriptor is suggested to characterize the3D graphical representation of aprotein, and a distance between two3D graphical representations is introduced tocompare the similarities of two corresponding proteins. The similarities/dissimilaritiesof the ND5protein sequences of nine different species are employed to illustrate theutility of our method. Furthermore, a correlation and significance analysis is providedin order to compare our results on similarities/dissimilarities and some other graphicalrepresentation results with the Clustal W similarity/dissimilarity results. Thecomparison results show that our approach has better correlations with Clustal W forall nine species than other approaches. In addition, the beta-globin proteins of50species are compared by our newmethod. Their distance matrix is calculated and the phylogenetic tree of50species isconstructed based on the neighbor-joining (NJ) method in PHYLIP package，whichis consistent with the taxonomic results.