Developing new attribute coding methods to prediction of disordered protein regions on based chaotic and physicochemical properties

Proteins located in all biochemical reactions have vital importance for continue of organism life. Physicochemical properties of amino acids that constitute proteins are the most important to formation for three-dimensional structure of protein and binding orders of amino acids linked by peptide bonds. Moreover, this condition leads to the lowest energy level in the cell. In this way, protein chains are folded naturely and created proteins. Each of ordered proteins is called the conformation. And also, disordered proteins can be formed. Disordered proteins cause many metabolic disease in organisms such as cancer, alzheimer, cardiovascular, cystic fibrosis and diabetes. Therefore, prediction of disordered regions in proteins is a significant threshold in terms of development of treatment for the diseases. In the literature, many studies have been applied to the prediction of disordered regions in protein in vitro and in silico enviroments. In silico study performances on computer has some advantages like cost and time-consume relatively in-vitro studies. Prediction of disordered regions have been determining by various machine learning algorithms and bioinformatics methods in silico. In this thesis, primarily, protein sequences coded by physicochemical properties and examined in phase space. Moreover, physicochemical properties on based chaotic structure was determined. And than, it has been developed using machine learning algorithms determined physicochemical properties of the amino acids such as Feature Encoding Method by Selected Best Physicochemical Properties and Features Encoding Method by Lyapunov Exponents and Selected Best Physicochemical Properties. Proposed methods have been applied with classifier algorithms on DisProt and PDB datasets to the prediction of disordered region in proteins. According to obtained experimental results that our methods have demonstrated the highest performance than most methods in the literature.