The high rates of failure in oncology drug clinical trials highlight the problems of using pre-clinical data to predict the clinical effects of drugs.Here we present two methodology innovations on network pharmacology modeling.(1) We hypothesize that the gene network associated with cancer outcome in heterogeneous patient populations could serve as a reference for identifying drug effects.We proposed a novel in vivo genetic interaction between genes as synergistic outcome determination in a similar way to ynthetic lethality.We scanned above genetic interactions based on microarray profiling for cancer prognosis, and identified a cluster of important yet epigenetically regulated gene modules.By projecting drug sensitivity-associated genes on to this network, we could define a perturbation index for each drug based upon its characteristic perturbation pattern.Finally, by using this index, we significantly discriminated successful drugs from the candidate pool, and revealed the mechanisms of drug combinations.Thus, the prognosis-guided synergistic gene-gene interaction networks could serve as an efficient in silico tool for pre-clinical drug prioritization and rational design of combinatorial therapies.Part of this work was published, and we will present new results on this project.(2) MicroRNAs (miRNAs) play a key role in the regulation of the transcriptome and have been identified as a key mediator in human disease and drug response.we introduced a novel concept, the Context-specific MiRNA activity (CoMi activity), to reflect a miRNAs regulation effect on a context specific gene set.Using breast cancer as an example, we examined the CoMi activity based on a Gene Ontology (GO) term as context.Interestingly, we found that chemotherapeutic drug treatment can counteract the dis-regulated CoMi activity in the cancer-specific network.For instance, 100% of down-regulated CoMi activities in a " core" breast cancer network contains apoptosis-related GO terms that could be counteracted by Paclitaxel treatment.By defining a Stability Index for in silico drug screening, we found CoMi activity signatures strikingly outperformed the traditional CMAP method or mRNA-based signatures.Thus, the dynamic remodeling of context-specific miRNAs regulation network could reveal the hidden miRNAs that act as key mediators of drug action and facilitate in silico cancer drug screening.