Platform for ligand-based toxicity prediction

The in-silico prediction of compound toxicity is a very important task, i.e. for the reduction of animal testing (http://www.bb3r.de/). Thus, we will implement Python scripts to set up a machine learning model for toxicity predictions. The data from freely available sources for toxicity predictions will be collected. This test and training data together with different molecular descriptors will be the basis for our toxicity predictor. Besides exhaustive evaluation of the models, a new method will be developed to extract the molecular features responsible for the toxic effect.

If successful, the model will be made available as a user-friendly web-server to allow other researchers to predict the toxicity of their compound of interest. A potential publication would be the ultimate goal of our work.

Used programming languages: Python (scikit-learn, RDKit), Webserver (django, json, jquery, …)

For further information on the research interests of the group see https://physiologie-cbf.charite.de/en/institute/workgroups/team_volkamer