SVM Model to Predict Human Death Domain Protein-Protein Interactions Based on Amino Acid Composition

Protein-Protein Interactions (PPIs) play crucial role in regulation of virtually all biological processes in any living system such as DNA transcription, replication, metabolic cycles and signaling cascades. The PPIs also play an important role in the complex process of cell death which occurs via apoptosis and necrosis in eukaryotic cells. The PPIs detection via high throughput experimental methods are time consuming, expensive and are generating huge amount of PPIs data. Therefore, there is need to develop computational methods to efficiently and accurately predict PPIs. This study attempts to develop computational model for predicting human death domain PPIs. First, the protein primary sequences are encoded into descriptors based on amino acid composition of proteins which are monomers of protein. Then, the support vector machine and sequential minimal optimization of WEKA tool is employed to classify interacting and non interacting protein pairs. The various kernel functions were evaluated to build the model and it is observed that libSVM with linear kernel is found to be the best on the basis of performance measures. The validation has been performed by 10 fold cross validation technique. The optimum model gives us the accuracy of 76.47% in predicting human death domain protein-protein interactions. Such models can be useful in providing PPI information of death domain proteins which can be useful in understanding the molecular mechanisms involved in death of cells taking place due to ageing, programmed cell death and various diseases.