With recently developed high-throughput technologies that allow us to gather biomedical data on genome-wide scale under a wide range of experimental conditions, scientific discovery has shifted from labor-intensive to computationally intensive task. The project will develop and apply a set of computational tools for inference of the mechanism of action of pharmacologically active substances in a model organism S. cerevisiae. In the application we will use a set of chemical-genomics profiles, that is, currently the most informative source on the interactions between drugs and genes. Data mining will be used to uncover the mechanism of drugs action. We will combine data analysis with in silico experiment planning techniques, and carry the proposals out on a robotic platform to increase the reliability of proposed hypotheses.

The expected principal results of this projects are A) bioinformatics toolbox (data analysis through clustering and classification of complex, genome-wide profiles, experiment planning through active learning), B) identification of a set of marker genes/mutants with high information content to predict the mechanism of drugs action, and C) a prototype of a high-throughput experimental platform combining state-of-the-art technologies from molecular biology, laboratory robotics and computational analysis for rapid classification of molecules based on their chemical-genetic interactions. Chemical-genomics is a very young and promising field of functional genomics, requiring dedicated computational tools for their application. With current practical demonstrations in this field being presently at best rare, development and application of proposed knowledge technology tools for analysis and proposal of experiments in drug discovery should be regarded as highly original.