In addition, as drugs can act on multiple targets, secondary targets can be utilized for novel drug indications as well. Several systematic approaches of finding new uses for old drugs have been proposed. These methods can be broadly classified into two categories: target discovery based on chemical compound similarity and literature-based discovery. Compound similarity has been a popular approach to identify drug targets for drug repurposing. The assumption is that similar drug compounds have similar targets so that targets that are not shared between a pair of similar compounds can be identified as novel targets to the other. By identifying new targets for existing compounds, new drug indications can then be proposed. On the other hand, typical text mining methods focus on the extraction of knowledge such as protein-protein interactions from biomedical literature. These text mining efforts including the BioCreAtIvE challenge, a community effort that aims to advance the development of biological knowledge extraction systems, focus on the extraction of biological knowledge that is explicitly stated in the literature. Literature-based discovery methods go a step further by identifying EX 527 relevant knowledge through text mining so that new knowledge can be inferred from existing knowledge. Swanson’s ABC Model is a popular literature-based discovery methodology that was proposed to link two concepts through a commonly shared concept. Scientific concepts A and C form a relationship when concept A cooccurs with concept B in one publication while concepts B and C cooccur in another publication. Variations of Swanson’s ABC models have been described in the literature for the identification of indirect relationships. However, approaches based on cooccurrences of concepts within abstracts tend to generate too many hypotheses. Another direction for network-based approaches aims to uncover knowledge through the creation of biological networks. STITCH and ChemProt are examples of network-based approaches that take interactions extracted from literature and integrates with data from biological knowledge bases to create chemical compound-protein interaction networks. This kind of approach in linking the concepts does not consider the inherent relationships between the pairs of concepts such as interaction type and directionality of interactions, thus leading to a large number of hypotheses. To handle large networks that are generated by means of literature mining and other data sources, visualization tools have been proposed to assist the discovery of novel drug indications. In this paper, we propose a new literature-based discovery approach for drug repurposing that integrates facts from various sources to infer novel indications by means of automated reasoning. Our approach captures the various effects of drugtarget interactions inside cells as well as the molecular mechanisms of diseases. Using cancer as an example, we utilized the wealth of knowledge about cancer and encoded oncogenes and tumor suppressors as well as cancer-related biological processes as the domain knowledge for our method. Together with the proteinprotein interactions and gene-disease associations acquired from the literature, our approach identified drugs that are potential candidates for the treatment of cancer. By considering the interaction types and their directionality and the domain knowledge involved in the mechanism of action of drugs, our approach aims to produce biologically meaningful hypotheses for novel drug indications and can significantly reduce the number of hypotheses as compared to previous text mining and literaturebased discovery approaches.