In silico network-based methods have been developed for reducing costs of drug development. Biological networks (bio-networks) consist of biological associations and are heterogeneous depending on different biological contexts. Here, we use context-based bio-networks for predicting effects of drugs in the human body. We predict drugs having opposite effects on disease genes as treatments of the diseases with ‘disease context’ and ‘effect type’ information. Also, we reconstruct an anatomical context-specific network including ‘multi-level entities’ such as genes, biological processes, and diseases. Our constructed network includes intercellular associations. We employ the network for analyzing effects of drugs on diseases.