Karl Marti Toots defended doctoral thesis, that helps to better understand and evaluate the behavior of organic solutes in ionic liquids
Uko Maran
The doctoral thesis defended at the Department of Chemistry of the University of Tartu studied the distribution of organic solutes in a gas-ionic liquid environment using cheminformatics and machine learning methods. For the first time, using computational models, it was shown that the gas-ionic liquid partition coefficient of a solute depends on the ionic parts of the ionic liquid.Furthermore, in the case of bigger data, the inclusion of multiple structural components in the computational model helps to significantly improve the predictive power of machine learning models.
The application of artificial intelligence and machine learning (ML) in quantitative structure-property relationships (QSPR) enables in silico studies to understand the physicochemical properties of materials and chemical compounds. One such class of chemical compounds is ionic liquids, where understanding and evaluating the distribution properties of organic solutes provides a basis for the study and development of various applied chemical environments. As environmentally friendly alternatives to classical organic solvents, ionic liquids are an important research object. However, systematic studies of the structure-property dependence of the partition properties of ionic liquids are missing. This gap is filled by the present study.
The doctoral thesis aimed to investigate the relationships between the gas-ionic liquid partition coefficient (log K) and the structure of organic solutes and/or ionic components of the ionic liquid, using cheminformatics approaches.
“The research involved the use of theoretical molecular features and advanced ML methods to model the interaction mechanisms based on the structure of the solute and the structure of the ionic liquid in multi-component systems,” explained Karl Marti Toots, the author of the doctoral thesis.
According to the opponent of the doctoral thesis, Dr. Igor Tetko, head of the chemical informatics group at the Helmholtz Institute for Structural Biology in Munich, the study showed that it is important to include as much data as possible to improve the accuracy of models
Modeling and analysis of data sets corresponding to the structures of organic solutes, cations, and anions showed that the ML methods of random forest regression, support vector regression, and Gaussian process regression represent the dependencies between the solute and the ionic liquid encoded in molecular descriptors more effectively than conventional multilinear regression. At the same time, the latter is easier to interpret. Both linear and nonlinear models emphasize the critical impact of the composition of cations and anions on the distribution of the solute and its modeling.
According to the doctoral thesis supervisor, Uko Maran, Professor of Molecular Technology at the University of Tartu, scientists have always relied on experiments and data, i.e., they are data-driven, in the study of chemical compounds, in the explanation and evaluation of their properties and in the development of new compounds. "To facilitate such knowledge acquisition, machine learning and artificial intelligence have been tools in everyday use, which are currently being particularly sought after, and which Karl Marti has skillfully used in his research," he explained.
The doctoral thesis co-supervisor, Associate Professor Sulev Sild, added that the results of the doctoral thesis also show that modeling the entire solute-ionic liquid system, combining solute, cation, and anion features, improves predictive power for large and chemically diverse data sets, emphasizing the importance of multi-component approaches.
The other doctoral thesis co-supervisor, Associate Professor Jaan Leis, highlighted as a significant result that the molecular descriptors included in the derived models explain possible interactions between components, based on the classification of solute-solvent interactions.
The doctoral thesis author, Karl Marti Toots, further emphasized that in addition to mechanistic knowledge, the derived relationships allow the design of more selective and efficient ionic liquids for targeted industrial, environmental, and scientific applications.
The public defense of the doctoral thesis took place on September 5, 2025.
Photo, Author: Romet Peedumäe
