Predicting Critical Properties and Acentric Factors of Fluids Using Multitask Machine Learning

Knowledge of critical properties, such as critical temperature,pressure, density, as well as acentric factor, is essential to calculatethermo-physical properties of chemical compounds. Experiments to determinecritical properties and acentric factors are expensive and time intensive;therefore, we developed a machine learning (ML) model that can predictthese molecular properties given the SMILES representation of a chemicalspecies. We explored directed message passing neural network (D-MPNN)and graph attention network as ML architecture choices. Additionally,we investigated featurization with additional atomic and molecularfeatures, multitask training, and pretraining using estimated datato optimize model performance. Our final model utilizes a D-MPNN layerto learn the molecular representation and is supplemented by Abrahamparameters. A multitask training scheme was used to train a singlemodel to predict all the critical properties and acentric factorsalong with boiling point, melting point, enthalpy of vaporization,and enthalpy of fusion. The model was evaluated on both random andscaffold splits where it shows state-of-the-art accuracies. The extensivedata set of critical properties and acentric factors contains 1144chemical compounds and is made available in the public domain togetherwith the source code that can be used for further exploration.

Automated summary

This study developed a machine learning model that can predict critical properties and acentric factors of chemical compounds based on their SMILES representation, replacing expensive and time-consuming experiments. The model uses D-MPNN and graph attention network architectures, and incorporates additional atomic and molecular features, multitask training, and pretraining to optimize performance. It achieves state-of-the-art accuracies on both random and scaffold splits, and the dataset of critical properties and acentric factors is now publicly available along with the source code.