Secondary structure prediction of hemeglobin by using combined neural networks

ABSTRACT MSc THESIS SECONDARY STRUCTURE PREDICTION OF HEMOGLOBIN BY USING COMBINED NEURAL NETWORKS îremERSÖZ DEPARTMENT OF ELECTRICAL AND ELECTRONICS ENGINEERING INSTITUTE OF NATURAL AND APPLIED SCIENCES UNIVERSITY OF ÇUKUROVA Supervisor : Asst. Prof. Dr. Turgay ÎBRÎKÇİ Year: 2003 Pages: 60 Jury : Asst. Prof. Dr. Turgay ÎBRÎKÇİ Prof Dr. Seyhan TÜKEL Asst. Prof. Dr. Ulus ÇEVİK Proteins are one of the most important parts of an organism because of its vital important tasks. In this respect, to understand the life process of an organism, it is necessary to first know the protein's structure that is closely related to its function. Protein structures are described through four main hierarchical levels; Primary, Secondary, Tertiary, Quaternary. The primary structure is simply the sequence of amino acids, the secondary structure refers to the local conformations of the polypeptide chains, the tertiary structure describes how the secondary structure elements are arranged to form the overall shape of the chains and the interactions between one or more polypeptide chains gives the quaternary structure. Artificial Neural Networks are useful toolbox for secondary structures prediction of proteins. In this thesis a generalized regression neural network (GRNN), probabilistic neural network (PNN) and backpropagation algorithm (BP) were applied to the hemoglobin primary structure with different window sizes of amino acid sequences to predict the secondary structure that has helix and coil. Then all results of the networks are combined with GRNN. The data set is prepared with 20 alpha- 141 and 20 beta- 146 hemoglobin chains from Protein Data Bank. The overall success rate of GRNN is around 90.2-91.7% for beta chains and 85.9-87.3% for alpha chains. The PNN achieved between 91.2-92% overall accuracy for beta and 85.4-86.5% for alpha. BP has the overall success rate of 89.9-92.5% for beta, 86.5-90.3% for alpha. There is no significant improvement in prediction accuracy with CNN. Keywords: Generalized Regression Neural Network, Probabilistic Neural Network, Backpropagation, Combined Neural Networks, Hemoglobin. II