ToxicBlend: Virtual Screening of Toxic Compounds with Ensemble Predictors
This work addresses the problem of timely toxicity assessment for pharmaceutical researchers, offering a practical tool to reduce compound discards, though it is incremental in its method combination.
The paper tackled the challenge of predicting compound toxicity in pharmaceutical screening by proposing an ensemble machine learning approach that combines extreme gradient boosting with neural networks, achieving significant performance improvements over existing state-of-the-art models on ToxCast and Tox21 datasets.
Timely assessment of compound toxicity is one of the biggest challenges facing the pharmaceutical industry today. A significant proportion of compounds identified as potential leads are ultimately discarded due to the toxicity they induce. In this paper, we propose a novel machine learning approach for the prediction of molecular activity on ToxCast targets. We combine extreme gradient boosting with fully-connected and graph-convolutional neural network architectures trained on QSAR physical molecular property descriptors, PubChem molecular fingerprints, and SMILES sequences. Our ensemble predictor leverages the strengths of each individual technique, significantly outperforming existing state-of-the art models on the ToxCast and Tox21 toxicity-prediction datasets. We provide free access to molecule toxicity prediction using our model at http://www.owkin.com/toxicblend.