Simultaneous inference of cell-line-specific gene regulatory networks and mode-of-action of drugs from drug-induced gene expression measurements

Personalized molecular networks help in elucidating drug resistance mechanisms, finding new therapeutictargets, and predicting the effectiveness of combination therapy. In this study a method for simultaneouslyinferring gene regulatory network and mode-of-action of drugs from gene expression measurements of drugperturbedcancer cell-lines is proposed. An inferred model is a dynamic state-space representation whosevariables are genes and inputs are drug dosages. Such a model includes gene-gene and drug-gene (mode-ofaction)relationships. To implement the proposed method, drug-induced expression levels of nearly ۱۰۰۰landmark genes for a set of cell-lines in the LINCS database were employed. Since no time-series geneexpression data are available the cell-lines are assumed to be in steady-state. The central challenge of thepresent study is under-determinedness for which a set of solutions are examined. Finally, the performance ofthe proposed method is evaluated by performing a cross-validation on the set of drugs for a specific cell-line.

Keywords:

Dynamical system , State-space model , Network inference , Gene regulatory network , Drug mode-of-action , Under-determinedness , Cancer cell-line