A NOVEL APPROACH TO PREDICT PROTEIN-PROTEIN INTERACTION AND FUNCTION

Abstract

Proteins interact with each other for a common purpose such as cell formation, metabolic cycles and numbers of other. They come in contact with each other following the chemical properties. The understanding of protein-protein interactions, in metabolic networks is an important aspect of molecular biology and biochemistry. The prediction of protein interactions among the proteins is important as it has been seen that interacting proteins serve a common purpose. Thus an unknown protein can be annotated knowing its interacting partners. It makes two problems for the same reason. The first one is to- predict the protein interactions among the proteins and second one is to predict the functions of an unannotated protein.. The prediction system uses protein sequence data.. to make the system learned. Based on sequence data the information about the structure of protein and the type of forces that mediate. the interaction can be extracted. This solution to this problem has a great help in the field of drug design. The need arises for the computational techniques due to the fact that the proteins data is available in a huge amount and biochemical research on each protein is too costly. The dissertation presents a new, approach, which takes into consideration the structure of proteins as well the types of forces that mediate the interactions among the proteins. The heuristic technique has been applied to find the interface surface which. is active in a protein during interaction. It makes the framework for learning the prediction system. Since finding an exact interface that takes part in the interaction is practically complicated, an optimum interface can always be found. The techniques applied such that it takes into consideration of each minute part of the protein sequence data., The two problems are dealt one by one. First the prediction system is designed which' predict 'the interaction and 'the approach is applied to the function.. prediction problem taking the produced transitional results of interaction problem as input to the function prediction problem. •The system modeled in this dissertation gives 74% accuracy, of prediction. The approach is implemented in C language under Windows XP'platform'. The dataset used for the implementation is collected from DIP (database of interacting proteins). Ill