Imputation of sensory properties using deep learning – ACS Spring 22
Presented by Dmitriy Chekmarev (IFF) and Samar Mahmoud (Optibrium), on 20 March 2022 at the ACS National Meeting and Exposition, USA
We demonstrate how Augmented Chemistry®, a unique deep learning method, can learn from higher throughput data together with limited panel data to provide high-quality imputations for sensory properties.
• Find novel active compounds with 3D virtual screening against known actives:
• Demonstrate the imputation of sparse physicochemical and sensory data
• Compare the results with conventional quantitative structure-activity relationship (QSAR) methods
• Present robust uncertainty estimates generated by the imputation model, highlighting the most accurate predictions to guide decision making
• Show that imputation more accurately predicts activity cliffs, where small changes in compound structure result in large changes in sensory properties.
You can download the presentation slides as a PDF