PreSPI+

Abstract

Many techniques have been developed in various different ways to computationally predict protein-protein interactions. However their prediction accuracies are not good enough to be serviced to general users. In this paper, we develop a domain combination based protein-protein interaction prediction method and service system named $preSPI^{+}$. The developed prediction method shows superior prediction accuracy to conventional predictions methods. High sensitivity (77%) and specificity (95%) are achieved for test protein pairs containing common domains with learning sets of protein pairs in Yeast. The stability of the method is also manifested through the test over DTP CORE, HMS-PCI, and TAP data. Inter-species validation of the method is also performed to test the extensibility of the method. Good prediction accuracies are obtained if there exist overlapping domains between learning and testing sets of protein pairs. This implies our method can be used for any species, once enough protein interaction data is a accumulated in the Internet. Performance, openness, and flexibility are adopted as design goals of $preSPI^{+}$ and each design goal is successfully achieved by processing time taking services in parallel, and using Web Services standards and layered architecture respectively. In this paper, the functions of $preSPI^{+}$ are briefly introduced and several services`` usage guides are explained in comprehensive manner with their graphic user interfaces.