Congrats to Xinpo on his ACS Omega paper published on Aug 12 comparing machine learning and deep graph learning for chemical ecotoxicity prediction. The study highlights the superior performance of GCN for environmental safety modeling, especially for same-species predictions. However, cross-species prediction remains challenging. Immediate testing of chemicals is available at https://app.cbbio.online/ecotoxicology/home. #GCN #MachineLearning #Ecotoxicology #ACSomega

The study employed molecular fingerprints, embeddings, and graphs combined with multiple machine learning and graph neural network models to construct and evaluate chemical ecotoxicological predictions.

Visualization of attention weights from the GCN model (trained on the F2F dataset) for four chemical compounds. (A) Sodium pentachlorophenate (toxic); (B) n-Octylphenol (toxic); (C) 4-chloro-3-tert-butylphenyl cyano dimethoxyphosphonate (less toxic); and (D) 2,4,6-trinitrophenol (less toxic).

For more details, please refer to the full paper.