Improving protein protein interaction prediction by choosing appropriate physiochemical properties of amino acids

Protein-protein interactions refer to the physical contact established between two proteins. They occur where two proteins bind together, for some biological function. Interacting protein sets obtained from Database of Interacting proteins (DIP) were analyzed and the data is generated by considering physicochemical properties and High Quality Indices (HQI) that is HQI-8, HQI-24 and HQI-40. In our research we analyze the data sets obtained by considering HQI-8, HQI-24, and HQI-40, and compare the prediction accuracy results of the respective data sets, and determine the set of features that is most suitable for prediction. Prediction is done with the help of Support Vector Machines (SVM). According to our results we conclude that the data set containing the unigram, bigram, and HQI-40 properties, showed maximum accuracy in prediction results. Protein-protein interactions are important for majority of biological functions, for example, the signals from the exterior of a cell are sent to the interior of the cell, using protein-protein interactions of signaling molecules. This process is signal transduction, and it plays an important part in many biological processes, and diseases like cancer.