Utilizing Machine Learning to Predict Physicochemical Properties of Unknown Chemical Agents
Abstract:
As the threat of terrorist attacks involving unknown chemical agents such as Novichok increases, understanding key physicochemical properties like vapor pressure and toxicity becomes critical for effective response strategies. In this presentation, we showcase the development of machine learning (ML) models designed to predict the vapor pressure of chemical threats, aiding in the timely management of escape and decontamination procedures in case of an accidental release. Additionally, we introduce an ML classification model tailored to organophosphorus compounds for toxicity prediction. Notably, our optimized ML model has been successfully applied to forecast the toxicity of chemical agents listed in the Chemical Weapons Convention (CWC). Furthermore, we developed a model to predict other essential properties, including hydrolysis rate and explosive potential, thereby contributing to broader preparedness for handling chemical threats.