Using Transcriptomics Data to Understand Drug Mode-of-Action

Understanding the mode of action of compounds is crucial during drug discovery and development, both for design and optimization purposes, but also, for instance, to anticipate side effects and for later approval. Recently, -omics data, and here in particular transcriptomics data, has been generated after drug treatments of cellular systems on a sufficiently large scale (for >20,000 compounds) that allows us to understand connections between drug-target interactions responsible for drug action and gene expression changes measured downstream systematically.

Hence, this work will greatly enhance our understanding of links between ligand-target interaction and transcriptomics changes on the fundamental level, and allow us to utilize a convenient type of readout for systems-wide mode-of-action analysis on the practical level. This project will in particular develop new and evaluate existing causal reasoning and related network approaches to predict the most relevant targets to a given differential gene expression change. To this end, we will employ the ConnectivityMap and LINCS databases which link compound structures to transcriptomic changes on a large scale (>20,000 compounds for LINCS), and link those changes to the known protein targets of drugs (obtained from databases such as DrugBank and ChEMBL) via mechanistic modelling using systems biology information on cellular signaling and other relevant information.

Subsequently, once those relationships have been sufficiently well understood, we will define prospective experiments in cellular systems, such as compounds selected using the developed algorithms which will be screened in microglial cell lines in collaboration with Eli Lilly. Changes in gene expression will be measured and compared to prediction, and phenotypic responses will be measured and evaluated with respect to agreement with the model predictions.