Fundamental Study of Solvated Crystal Structures Using Data-Mining and Machine-Learning Method

For solvated crystals, the implications in material synthesis, molecular biology, environmental science (e.g. CO₂ capture, metallurgy) and especially the pharmaceutical industry are enormous due to the effects in physicochemical properties of materials which in turn can influence the pharmaceutical synthesis steps during drugs manufacturing. For the formation of solvated crystals, the fundamental understanding remains elusive. To obtain an in-depth systematic study, a high-throughput screening of data-mining study of large dataset of solvated crystals might be useful. To achieve this goal, a dataset consists of organic, ionic and nonpolymeric molecules was extracted from the Cambridge Structural Database. In this study, we will share with you our findings in one particular types of solvates, i.e. dimethyl sulfoxide solvates based on several machine-learning models (e.g. random forest, gradient boosting, etc.). Based on or current models, the prediction of the behavior of ~ 80% of the data correctly using machine-learning model is attainable.