Towards Greener Solvents: A Machine Learning (ML) Pipeline for Sustainable Alternatives

Organic solvents are widely used in industry but pose significant environmental and health risks, with many classified as ecotoxic or carcinogenic. Solvent selection guides (SSGs) were developed as a tool to help scientists responsibly and safely select solvents. However, current SSGs are limited in scope and rely heavily on experimental data, making the process of finding “greener” solvent alternatives slow and costly. In this work, we present a computational pipeline that 1) predicts the "greenness" of any solvent and 2) suggests potential replacements. Trained on the GlaxoSmithKline Solvent Sustainability Guide–which scores solvents according to their environment, health, safety, and disposal impact–our Gaussian Process Regression (GPR) model can predict the greenness score of any solvent using only its SMILES. Incorporating Hansen Solubility Parameters (HSP), the pipeline can suggest replacements that are both greener and “similar” to the target solvent. This method provides a rapid, cost-effective approach for discovering greener solvents, with potential applications in green polymer synthesis and sustainable chemical process design.