Abstract Therapeutic and biomarker discovery requires comprehensive knowledge integration of disease, biological process, molecular component (i.e., DNA, mRNA, protein, microRNA, metabolite etc.), model system (i.e. cell line, animal, tissue, organ, etc.), drug activity, and clinical outcome, etc. Taking on the challenge, we are developing an in silico discovery engine which can efficiently manage all above aspects through a robust web-based interface and backend database. The in silico discovery engine enables access to information from diverse sources including biomedical literature, high-throughput gene expression analysis, translational research, and clinical trial results. Key components of the system comprise modules of knowledge capture, database, visualization, simulation, and hypothesis generation. Information retrieval procedures and tools were developed to capture biomarker data in all possible aspects including disease profile, drug action, molecular interaction and clinical outcome. Text-mining techniques were employed as initial filters to select publications for expert review. Information extracted from biomedical literature, public domain databases and proprietary sources was further standardized with our in-house ontology and controlled vocabularies. Lastly, interactions originally presented in natural language were translated into a well-structured format according to specific semantic rules designed specifically for computational purpose. The true power of this in silico discovery engine resides on its ability to uncover hidden relationships and generate de novo hypotheses through comprehensive modeling of disease, biological process and drug action. Particularly, we focused on developing molecular networks comprising drugs, gene products and microRNAs. The system captured information on more than 700 compounds, 3,000 genes and 1,000 microRNAs. Through presenting information in an integrated and computation-enabled format, the discovery engine enables investigation of molecular components in selecting indication, combating drug resistance and developing drug combinations. So far, we have developed integrated knowledge management solutions for 1) triple-negative breast cancers (TNBCs) with 333 compounds, 733 genes, and 43 microRNAs; 2) breast cancer stem cells (bCSCs) with 241 compounds, 496 genes, and 41 microRNAs; 3) epithelial-to-mesenchymal transition (EMT) with 180 compounds, 651 genes, and 127 microRNAs. Through modeling such drug-gene-microRNA networks, we were able to evaluate therapeutic targets which cannot be assessed by conventional methods, for example, the role of ERK in neurodegenerative diseases including Alzheimer's and Parkinson's. Similar exercises helped us to evaluate the effects of HDAC inhibitors on EMT, and assess combination effectiveness of 5-FU and cisplatin. The current study represents a paradigm shift in therapeutic and biomarker development, and demonstrates a fully integrated and knowledge-driven hypothesis-generation approach in drug discovery. Citation Format: Jian Zhu. Drug-gene-microRNA networks in therapeutic and biomarker development [abstract]. In: Proceedings of the AACR Special Conference on Chemical Systems Biology: Assembling and Interrogating Computational Models of the Cancer Cell by Chemical Perturbations; 2012 Jun 27-30; Boston, MA. Philadelphia (PA): AACR; Cancer Res 2012;72(13 Suppl):Abstract nr A30.