Abstract

The use of plastics has increased over the past decades due to their versatility and widespread use in packaging, construction, and agriculture. This ubiquity results in environmental and human health challenges because of the release of plastic monomers and additives. To tackle these problems, it is important to develop novel biodegradable plastics, aligned with frameworks like the Safe and Sustainable by Design (SSbD). The SOUL project is at the forefront of these innovations, as it aims to provide bio-based solutions to replace plastics in soil applications. As such, it represents an excellent testing ground for new tools to support SSbD assessment. Aligned with SSbD, the first step should be a hazard screening, with several cut-off criteria – Mutagenicity Category 1A or 1 B is amongst them. However, the process of compiling and checking mutagenicity evidence is time consuming, and data for novel chemicals is often not sufficient, hindering timely SSbD decisions. In this communication, it is presented an AI framework dedicated to evaluating the mutagenic hazard potential of plastic monomers. Using an empirically derived set of mutagenicity data, a machine learning predictive model was built that correctly classifies 92 % of plastic monomers as either Mutagenic Category 1A or 1B (H340) or Mutagenic in some capacity. The developed model enables rapid prediction of mutagenicity for any plastic monomer using only its SMILES representation, requiring no experimental input. By using open-source chemical descriptors it provides a reliable go/no-go classification for mutagenic potential for plastic development within the SSbD framework. This is a user-friendly tool for carrying out early-stage risk assessment that enables stakeholders to discover hazardous compounds before product lab testing. Specifically built for plastic applications, this model will generate new insights, allowing for future evaluations of alternative materials, supporting the transition to sustainable solutions.

Download the paper.