Identification of intracellular signal transduction pathways plays an important role in understanding the mechanisms of how cells respond to external stimuli. The availability of high throughput microarray expression data and accumulating knowledge of protein-protein interactions have provided us with useful information to infer condition-specific signal transduction pathways. We propose a novel method called Gibbs sampler to Infer Signal Transduction pathways (GIST) to search dys-regulated pathways from large-scale protein-protein interaction networks. GIST incorporates different knowledge sources to extract paths that are highly associated with biological phenotypes or clinical information. One of the most attractive features of GIST is that the algorithm will not only provide the single optimal path according to the defined cost function but also reveal multiple suboptimal paths as alternative solutions, which can be utilized to study the pathway crosstalk. As a proof-of-concept, we test our GIST algorithm on yeast PPI networks and the identified MAPK signaling pathways are well supported by existing biological knowledge. We also apply the GIST algorithm onto a breast cancer patient dataset to show its feasibility of identifying potential pathways for further biological validation.