Abstract
Hepatitis B virus (HBV) is a major human pathogen which will lead to human morbidity and mortality. To investigate the mechanism underlying the pathogenic process of HBV, we have developed a method to predict interactions between HBV proteins and human proteins. Totally, 382 protein-protein interactions (PPIs) were identified from the results of interologs and support vector machine (SVM) method. In interologs method, we defined a HBV protein interacted with a human protein when its partner in known human PPIs is homologous with the HBV protein. Furthermore, we adopted SVM with conjoint triad feature representation which showed good performance to deduce PPIs. We evaluated our predictions by the gene set enrichment analysis on targeted human proteins of HBx (one of HBV proteins). The predicted targeted proteins tend to share similar biological functions with experimental ones and also HBx. The results suggest that the predicted PPIs are suitable for further verification to facilitate the understanding about pathogenesis of HBV.