The Polymer Society of Korea

Abstract

In response to growing chemical terrorism threats and the emergence of unknown hazardous substances like structurally diverse Novichok agents, this study presents advanced computational approaches to enhance chemical safety. The diversity of these agents poses challenges for detection using conventional techniques.
To overcome this, we developed classical machine learning models to predict key properties such as vapor pressure and toxicity, enabling faster risk assessment. Additionally, we explored quantum machine learning (QML) methods, including techniques like Data Re-uploading and Identity Block structures. These QML models, especially when combined, showed superior performance over classical neural networks in toxicity classification tasks.
The results demonstrate the promise of quantum-enhanced models in accurately assessing the risks of novel chemical agents, contributing to improved preparedness against emerging threats.