Modeling from Theory and Modeling from Data: Complementary or Alternative Approaches? The Case of Ionic Liquids

Area of Science:

• Computational Chemistry
• Materials Science
• Cheminformatics

Background:

• Quantitative structure-property relationship (QSPR) models are crucial for predicting chemical properties.
• Traditional QSPR models for ionic liquids (ILs) often use unique, nonlinear models for broad data fitting.
• This approach can limit model applicability and interpretability.

Purpose of the Study:

• To develop novel chemoinformatic and chemometric data-driven QSPR soft models for ILs.
• To enable accurate prediction of key physicochemical properties of ILs.
• To provide accessible modeling tools for industrial research and development.

Main Methods:

• Utilized chemoinformatic and chemometric data-driven procedures.
• Developed QSPR soft models with local validity.
• Employed readily available and interpretable VolSurf+ descriptors.

Main Results:

• Successfully predicted physicochemical properties of ILs, including viscosity, density, decomposition temperature, and conductivity.
• Achieved local validity in QSPR models, enhancing predictive accuracy for specific IL subsets.
• Demonstrated the utility of VolSurf+ descriptors in developing effective IL property prediction models.

Conclusions:

• Data-driven QSPR soft models offer a valuable alternative to traditional approaches for IL property prediction.
• These models provide experimentalists with practical tools for IL design and optimization in R&D.
• The use of VolSurf+ descriptors facilitates the interpretation and application of the developed models.