Ionic Liquid Toxicity Prediction Using Deep Learning

Abstract

Every day we are exposed to different chemicals through food additives, cleaning products, cosmetics, and medicines - some of which may be toxic. However, testing the toxicity of all existing compounds through biological experiments is not financially or logistically feasible. So we launched this new project in the field. Within the "Toxicology in the 21st Century" initiative. The objective of this project is to assess the performance of computational methods in predicting the toxicity of chemical compounds. The latest toxicity forecasting methods rely on specially designed chemical prescriptions developed over decades. Although deep learning is new in this area and has not been applied to toxicity prediction before, it outperformed all other participatory methods. In this application note, we show that deep neural networks automatically learn all the features of compounds up to their toxicity ratio