Multi-Label-Classification for REACH

This web page presents (Q)SAR models to predict the toxic effect of chemical compounds. The available (Q)SARs are Multi-Label-Classification models, they predict multiple endpoints simultaneously for each compound.

Background

The service was created within the project 'Strategies to develop chemical categories in the context of REACH' (0315546B), founded by the Federal Ministry of Education and Research ( BMBF ). A list of participating project partners can be found here . This work was presented with a poster at the DGPT 2014 , a publication is in the making.

List of MLC-Models

• Carcinogenicity prediction with Ensemble of Classifier Chains ― proof of concept on a well known dataset
• Rat toxicity prediction with Ensemble of Classifier Chains ― most predictive rat toxicity model
• Rat toxicity prediction with Predictive Clustering Trees ― less predictive, provides categories