پيش بيني نقشه تماس پروتئين توسط روش ماشين گروهي

چكيده انگليسي :

105 Protein Contact Map Prediction Using Committee Mahine Approach Narjes Khatoon Habibi nhabibi@ec iut ac ir Date of Submission April 2009 Department of Electrical and Computer Engineering Isfahan University of Technology Isfahan 84156 83111 Iran Degree M Sc Language FarsiMohammad H Saraee saraee@cc iut ac irAbstractBioinformatics is a multi disciplinary field that applies principles from Mathematics Physics Chemistry andComputer Science to widespread numerous and complex biological data The aim of bioinformatics is tosolve biological problems in molecular level Proteins are the basic functional units of living organism s cellswhich carry out almost all the activities of life Each protein molecule comprises of an amino acid chain Four structures are defined for protein namely primary secondary tertiary 3 D and quaternary Primarystructure is the amino acid chain Secondary structures are local structures which are generated by hydrogenbonds The most common types of the secondary structures are alpha helices and beta sheets Tertiarystructure is the final shape of a protein molecule which is formed by folding amino acid chain In fact it isthe spatial status of the secondary structures in relation to eachother The quaternary structure is formed bycollection of several proteins Protein function is dependent on its tertiary structure The determination of the3 D structure of proteins is an important step toward understanding the behavior of them Tertiary structureitself is dependent on amino acid chain The determination of tertiary structure is not as straightforward process as the primary one Existingexperimental processes to determine tertiary structure are costly and time consuming which encourageresearchers to find methods to predict protein tertiary structure only based on its amino acid chain Contactmap prediction is one of such method Protein contact map is a simplified 2 D representation of proteinspatial structure The purpose of contact map prediction problem is to compute an estimate of contact map ofa protein based on its primary structure and features that are computable or predictable from primarystructure Over the years a variety of statistical and machine learning methods have been developed topredict contact map Committee machine is a machine learning method which divides the learning task among a number oflearners and input spaces into some sub spaces Learner s responses to an input are combined to produce thesystem s final response that is more accurate than of every individual s response The aim of this research isto propose a novel method for contact map prediction based on committee machine In the proposed method learner group is a set of neural networks Different features are extracted and then in two phases the learnergroup is generated as predictive model The important principle in evaluating contact map prediction is theratio of correct predicted contacts to all predicted ones To analyze the results of the proposed model twoother methods are implemented and their results are compared and presented The results show considerablegain which is achievable by committee machine approach in contact map prediction problem Key Words Bioinformatics Machine Learning Committee Machine Neural Network Protein Contact Map Contact Map Prediction

حبيبي، نرجس خاتون

پيش بيني نقشه تماس پروتئين توسط روش ماشين گروهي

بيوانفورماتيك , يادگيري ماشين , شبكه عصبي