Prediction and Interpretation of Polymer Properties Using the Graph Convolutional Network

We present machine learning models for the prediction of thermal and mechanical properties of polymers based on the graph convolutional network (GCN). GCN-based models provide reliable prediction performances for the glass transition temperature (T-g), melting temperature (T-m), density (rho), and elastic modulus (SE) with substantial dependence on the dataset, which is the best for T-g (R-2 similar to 0.9) and worst for E (R-2 similar to 0.5). It is found that the GCN representations for polymers provide prediction performances of their properties comparable to the popular extended-connectivity circular fingerprint (ECFP) representation. Notably, the GCN combined with the neural network regression (GCN-NN) slightly outperforms the ECFP. It is investigated how the GCN captures important structural features of polymers to learn their properties. Using the dimensionality reduction, we demonstrate that the polymers are organized in the principal subspace of the GCN representation spaces with respect to the backbone rigidity. The organization in the representation space adaptively changes with the training and through the NN layers, which might facilitate a subsequent prediction of target properties based on the relationships between the structure and the property. The GCN models are found to provide an advantage to automatically extract a backbone rigidity, strongly correlated with T-g, as well as a potential transferability to predict other properties associated with a backbone rigidity. Our results indicate both the capability and limitations of the GCN in learning to describe polymer systems depending on the property.

Automated summary

This study presents machine learning models based on the graph convolutional network (GCN) for predicting the thermal and mechanical properties of polymers. The GCN-based models perform well in predicting properties such as glass transition temperature, melting temperature, density, and elastic modulus, with the best performance observed for the glass transition temperature and the worst for the elastic modulus. The study shows that the GCN representations of polymers can provide prediction performances comparable to the popular extended-connectivity circular fingerprint (ECFP) representation. The combination of GCN and neural network regression slightly outperforms ECFP. The study investigates how GCN captures important structural features of polymers to learn their properties and finds that the representation space organization adaptively changes with training and through the neural network layers, facilitating the prediction of target properties based on the structure-property relationships. The study also demonstrates that GCN models can automatically extract backbone rigidity, which is strongly correlated with the glass transition temperature, and have the potential to predict other properties associated with backbone rigidity.