Network Analysis of Drug Responses

The pharmaceutical industry is facing unprecedented pressure to increase its productivity. Attrition rates in the later stages of development have risen sharply, with toxicity and lack of efficacy being the main bottlenecks. To address both these safety- and efficacy-related issues, a better understanding of the complex biological response to drug treatment is vital. Although many drugs exert their therapeutic activities through the modulation of multiple targets, for most drug candidates these targets are largely unknown and identification among the thousands of gene products remains difficult. Furthermore, this polypharmacology often goes hand in hand with unintended off-target effects. A better knowledge about these drug-protein interactions, along with the molecular pathways involved and the associated diseases, could be of substantial value to drug development, in particular to predict side effects and explore potential drug repositioning. In this thesis we propose a computational method to support the identification of putative targets of a drug by means of a dual approach combining gene expression and protein interaction data with chemical structure similarity. The first component of our method tackles this target identification problem by the analysis of gene expression following drug treatment in the context of a functional protein association network. More specifically, genes are prioritized as potential targets based on the transcriptional response of functionally related genes. To this end, differential expression signals are diffused over the network either using a kernel-based random walk or on the basis of connectivity correlations between nodes. This diffusion strategy was evaluated on 235 publicly available gene expression datasets for treatment with bioactive molecules having a known target. AUC values up to 92\% indicate the predictive power of integrating experimental gene expression data with prior knowledge on protein interactions to identify the targets of a drug. As gene expression analysis is frequently employed to study the effects of drug compounds on cells, we have developed Galahad to facilitate the analysis of this particular data type and enable an in-depth exploration of a drug's mode of effect. Galahad is a web-based application for the analysis of gene expression data from drug treatment versus control experiments, aimed at predicting a drug's molecular targets and biological effects. Galahad provides data quality assessment and exploratory analysis, as well as computation of differential expression. Based on the obtained differential expression values, drug target prioritization and both pathway and disease enrichment can be computed and visualized. All of the above functionalities are demonstrated on gene expression data for treatment with well-characterized drugs. In addition to using gene expression, drug-protein interactions can also be predicted from structural information. Building on the similar property principle, the second component of our drug target prioritization method ranks proteins as potential drug targets based on the interaction with compounds structurally similar to the drug of interest. To this end compound-compound similarity scores are combined with compound-protein interaction scores. Both our structure-based and expression-based approach produce a genome-wide ranking of potential targets that can eventually be fused to obtain a single ranking. This combined approach has been evaluated on a test set of 62 small molecule drugs with known targets. AUC values of up to 95\% were obtained. These results indicate the predictive power of combining gene expression data and structural information for a drug of interest with known protein-protein and protein-compound interaction information respectively, to identify the targets of that drug. As such our method can aid in gaining a better knowledge of a candidate drug's mode of action and its off-target effects and thus be of value in the drug development process.status: publishe