Predicting protein-protein interactions based on protein-domain relationships

This paper proposes a new method that can predict the interactions between proteins intermediated by the proteindomain relations. We utilize the lazy expectation maximization (LEM) to compute an improved maximization likelihood estimation (MLE) model. The protein-domain relationships are extruded from Pfam database and the combined data set of Uetz and Ito are used as the source of protein-protein interactions. Finally, the efficiency and the effectiveness of our proposed approach can be validated by a better performance such as the sensitivity of 80.1%, the specificity of 43.5%, and the lesser computational cost.