VSEngine - Similarity & Property prediction based Virtual Screening Web Services
Academic scientists from the chemoinformatics group of Prof. Varnek (Laboratoire d'Infochimie, UMR 7177 - Université Louis Pasteur, Strasbourg) have developed over time various innovative virtual screening applications (novel - fragment & pharmacophore - descriptors, similarity scoring techniques, massively parallel mining for QSAR models, novel applicability domain definitions & linear/nonlinear consensus scoring schemes, condensed graph-based modeling of chemical reactivity, etc). One month ago, our group started an effort to set up a virtual screening web service allowing non expert users to access these techniques on line in order to mine for potentially interesting compounds within the web server's molecule database, ScreenDB. This latter is thought to regroup sets of commercially or academically available compounds, in addition of reference molecules of known properties - from publicly available QSAR training sets, intended to serve as "seeds" (specific alerts will be issued automatically as soon as a query is found to have returned an abnormally high concentration of reported actives against a given target). The challenges faced during the development of these integrated web services are of technical nature - making the various heterogeneous approaches to work with each other and with the classical substructure & similarity search engines. This can be quite naturally achieved with the ChemAxon toolbox, in charge of all the aspects concerning compound standardization & formal charge assignment, MySQL database management, molecular format interconversion, standard (sub)structure & similarity searches, physicochemical property & descriptor calculations, hit visualization, etc.