Academic

A two-year post-doctoral position in Computer Science and chemoinformatic is opened at the GREYC laboratory (CNRS UMR 6072) at University of Caen, www.greyc.unicaen.fr

Working context: the position has been accepted by the Emergence call for proposals funded by the Region Basse-Normandie.

Topic: Designing classifiers by using graph structures to predict toxicological properties of chemical compounds

The European Union has recently adopted a new regulation (the REACH regulation) about the production and commercialization of chemicals. The REACH regulation enables us to use alternate methods (typically in silico methods) in order to define the physicochemical profile of molecules.

A lot of research is devoted to the creation of classification rules to predict the activity of a molecule according to its structure. This post-doctorate position focuses on the processing of chemical information: its purpose is to contribute to the development of reliable decision rules and classifiers to assess the toxicity of a chemical compound according to its structure. In this context, this research includes an analysis of the relevance of two families of descriptors:

• The largest common subgraph. Molecular structures are modelized by graphs, hence the largest common subgraph to two graphs induces a notion of intersection between the molecular structures. However, there are several notions of largest common subgraph and there is a need to compare their relative interest of in the context of structure activity relationship.

• Kernel similarities. These methods are applied to graphs for few years. In spite of their implementation may be expensive, their performances are promising. We need to assess the relevance of kernel similarities in the prediction of chemical activity.

http://bingo2.greyc.fr/openPositions/postDocInnotox2GREYC_En