Signal transduction is important in many different aspects of cellular activity. Many computational methods have been generated in mining signal transduction networks with the increasing of high-throughput genomic and proteomic data. However, more effective means are still needed to understand the complex mechanisms of signaling pathways. In this paper, we have developed a computational approach for generating models of signal transduction networks. Networks are determined entirely by proteinprotein interaction data without prior knowledge of any pathway intermediates. This approach should enhance our ability tomodel signaling networks and to discover new components of known networks. The precision and recall values of ourmethod are comparablewith other existing methods. Our method is a more suitable method than existing methods for detecting underlying signaling pathways.