Computational methods are essential for evaluating the potential risks of compounds that lack experimental data, with skin sensitization being a critical endpoint in chemical safety assessments. The integration of knowledge derived from proprietary data sources into in silico models has previously been shown to increase the predictive accuracy and reliability of in silico models.
blank
Recently, Leadscope has collaborated with the industry to update our skin sensitization models. The effort included integrating knowledge derived from proprietary data sources into the knowledge base of the expert rules-based system. Additionally, several hundred new structures were added to the training and reference databases. Through a collaborative effort with industry partners, historical skin sensitization data were collected for 925 compounds assessed in the Local Lymph Node Assay (LLNA) or Guinea Pig Maximization Test (GPMT). The updated models demonstrated 82% accuracy in predicting strong/extreme sensitization outcomes in a proprietary dataset. This study supports the hypothesis that knowledge from corporate databases can advance in silico models with expanded predictive capabilities and performance.
We are excited to share the results of this collaborative effort in a symposium titled ‘Extraction of Knowledge from Proprietary Datasets for the Advancement of QSAR Models’ at the SOT 64th Annual Meeting and ToxExpo in Orlando, Florida (Wednesday, March 19th, 9.05 AM-9.35 AM, Room W203A).
Please contact us to speak to an expert or learn more about our solutions here.
