Conformation-Independent QSPR Approach for the Soil Sorption Coefficient of Heterogeneous Compounds

Area of Science:

• Environmental Chemistry
• Computational Chemistry
• Toxicology

Background:

• Accurate prediction of soil sorption coefficient (Koc) is vital for environmental fate and risk assessment of organic chemicals.
• Existing methods for Koc prediction can be limited by descriptor types and dataset size.

Purpose of the Study:

• To develop a robust Quantitative Structure-Property Relationship (QSPR) model for predicting the soil sorption coefficient (Koc) of diverse organic non-ionic compounds.
• To identify optimal molecular descriptors for Koc prediction using readily available software.

Main Methods:

• Utilized a dataset of 643 organic non-ionic compounds.
• Generated 17,538 conformation-independent molecular descriptors using PaDEL and EPI Suite.
• Employed linear regression with the Replacement Method for variable selection.
• Developed a three-descriptor QSPR model and a single-descriptor model using CORAL.

Main Results:

• The best three-descriptor QSPR model demonstrated acceptable predictive capability on a test set of 550 compounds.
• A single-descriptor model using CORAL also showed good predictive performance.
• The developed QSPR models compare favorably with previous approaches using different descriptor sets.

Conclusions:

• The study successfully established QSPR models for predicting soil sorption coefficients.
• The findings highlight the utility of conformation-independent descriptors and accessible software for environmental modeling.
• The developed models offer a reliable tool for estimating Koc values for environmental risk assessment.