AI for Materials Discovery: Biodegradable Polymers for Military Ration Packaging

The application of artificial intelligence (AI) methods in the discovery of biodegradable polymers is an increasingly important research direction due to the inefficiencies of traditionally used methods. The aim of this study is to explore and compare traditional machine learning (ML) models such as Random Forests (RF) and XGBoost to more recent GNNs (GNNs) to rapidly predict various properties applicable for biodegradable military food packaging applications from input polymer SMILES. After collecting applicable polymer datasets from the literature, we explore appropriate feature embeddings for representing polymers for use in our selected models. Model selection and analysis is conducted on the property of Glass Transition temperature (Tg), with our best model evaluated on additional properties. Despite our results being preliminary, they indicate that the models we experimented with displayed promising results for producing reliable polymer property predictions . We also carried out investigations into the efficacy of pre-training GNN models. Finally, the potential usefulness and limitations of applying ML tools to accelerate materials discovery is discussed.