Machine learning is being increasingly utilized in different research fields for its capability to gather vast quantities of data and then generate a series of prediction models for new input. Density and refractive index (nD) are two important properties related to the van der Waals energy of a molecule; thus, accurately predicting the two properties has a great value in both molecular mechanics force field development and solvation-free energy and solubility prediction of any arbitrary molecules.<br/><br/>In this study, we gathered molecule characteristics information of around 10,900 organic compounds and collected their density and refractive index values; subsequently, the distinct GAFF (General AMBER Force Field) and RDkit descriptors of the compounds were generated then applied to train various prediction models with a variety of machine learning algorithms for both properties. Very encouragingly, GAFF atom type-based and RDkit molecular descriptors yielded various robust models with low average percent errors (APE) and low Root means square errors (RMSE). Specifically, for density, the RMSE values are 0.021 and 0.020 for GAFF and RDkit descriptors, respectively. Note that this performance is much better than achieved by the state-of-the-art molecular dynamics simulations for most current small molecule force fields. As for the refractive index, again, the GAFF descriptor is marginally worse than RDKit, and the test set RMSE values are 0.025 and 0.021, respectively. Interestingly, we found that the consensus prediction using the best models of both descriptors can significantly reduce RMSE to 0.015 for density but not for nD, probably due to the different nature of the two properties. The successful prediction of the two key molecular properties paves the road towards accurately predicting the solubility of an arbitrary solute in an arbitrary solvent; an endeavor not only facilitates the pharmaceutical industry to develop better drug candidates but also increases efficiency regarding overall wet lab work.

Toward predicting solubility of arbitrary solutes in arbitrary solvents: Prediction of density and refractive index using machine learning algorithms

Where thousands of chemistry professionals meet to share ideas and advance scientific and technical knowledge. Spring 2023 abstracts, posters, and presentations will launch at the end of March.

ACS Spring 2023

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.