Abstract Solubility is a crucial property of organic compounds, impacting their potential applications in synthetic chemistry, materials science and drug design. Moreover, in technological processes mixtures of solvents are often utilized, making the solubility assessment more complicated. Predicting solubility values in mixtures of solvents from a molecular structure can help to address this issue, although a large and diverse dataset is needed to effectively pursue data-driven studies. In this research, we present a dataset containing 175166 experimental solubility values within the temperature range from 252 to 383 K for 810 organic compounds possessing 3001 unique solute-binary solvent systems and 750 unique binary solvent mixtures extracted from 1115 peer-reviewed articles. The solubility data and molecular structures of solutes and solvents are translated to a unified machine-readable format, facilitating data analysis and machine learning model development. An interactive online tool for visualization and navigation through the data has also been developed. This dataset can serve as a comprehensive benchmark for predicting solubility in mixtures of solvents.
Malikov et al. (2026) studied this question.