Toxicity of ionic liquids: database and prediction via quantitative structure-activity relationship method.

Area of Science:

• Environmental Chemistry
• Toxicology
• Computational Chemistry

Background:

• Ionic liquids (ILs) are versatile solvents with growing applications.
• Understanding ILs' toxicity is crucial for their safe development and use.
• Existing toxicity data is often fragmented, necessitating a centralized resource.

Purpose of the Study:

• To establish a comprehensive database of ionic liquid toxicity.
• To analyze the structure-toxicity relationships of ILs.
• To develop predictive Quantitative Structure-Activity Relationship (QSAR) models for IL toxicity.

Main Methods:

• Compilation of a large-scale IL toxicity database (>4000 data points).
• Qualitative analysis of structure-toxicity correlations.
• Development and comparison of Multiple Linear Regression (MLR) and Support Vector Machine (SVM) QSAR models.
• Selection of four key molecular parameters using the Heuristic Method (HM).

Main Results:

• QSAR models demonstrated high accuracy in predicting IL toxicities (EC50 values) against the IPC-81 cell line.
• The SVM model achieved superior predictive performance (R²=0.958, RMSE=0.234) compared to the MLR model (R²=0.892, RMSE=0.329).
• Increased relative oxygen atom content in IL molecules correlated with decreased toxicity.

Conclusions:

• The developed SVM-based QSAR model provides a reliable tool for predicting IL toxicity.
• Structural features, specifically oxygen content, significantly influence IL toxicity.
• The comprehensive database and predictive models facilitate safer design and application of ionic liquids.