Pathogen-Host protein-protein interaction prediction using Kernelized Matrix Factorization

Pathogens infect their host by exploiting host cellular mechanisms and evading host defense mechanisms through molecular pathogen-host interactions (PHIs). Discovering of interactions between pathogen and Human proteins may lead to detecting important interactions responsible for infectious diseases and delivering new drugs. In this paper, we propose to use a method based on Bayesian Matrix Factorization for predicting PHIs. Formulating the problem in this way relieves the need for negative samples. This is significant since there is no available dataset to introduce negative samples for most of the pathogenic systems. We present new features based on interaction pattern and feed them to a method in form of similarity matrices between proteins. Our experiments verify that, proposed approach outperforms state of the art methods recently proposed for PHI prediction.

Pathogens infect their host by exploiting host cellular mechanisms and evading host defense mechanisms through molecular pathogen-host interactions (PHIs). Discovering of interactions between pathogen and Human proteins may lead to detecting important interactions responsible for infectious diseases and delivering new drugs. In this paper, we propose to use a method based on Bayesian Matrix Factorization for predicting PHIs. Formulating the problem in this way relieves the need for negative samples. This is significant since there is no available dataset to introduce negative samples for most of the pathogenic systems. We present new features based on interaction pattern and feed them to a method in form of similarity matrices between proteins. Our experiments verify that, proposed approach outperforms state of the art methods recently proposed for PHI prediction.