Prediction of Bacteriophage Protein Locations Using Deep Neural Networks

In phage therapy, bacteriophage proteins are used to kill bacteria that cause infection. The knowledge of the location of the bacteriophage proteins plays an important role here. In this paper, we propose a supervised learning based method to predict the locations of bacteriophage proteins. First, we address the problem of predicting whether a bacteriophage is extracellular or located in the host cell. Second, we also address the subcellular location prediction problem of the phage proteins. For the host located proteins, the proteins could either be located in cell membrane or in the cytoplasm. We have successfully used deep feed-forward neural network on a standard training dataset and achieved good results for both of the prediction problems. Our method uses an optimal set of features for classification and achieves 87.7% and 98.5% accuracy for two of the prediction problems which is 3.5% and 6.3% improved than the previous state-of-the-art results achieved for these problems, respectively.