Quantitative structure-property relationships for the calculation of the soil adsorption coefficient using machine...

Area of Science:

• Environmental Chemistry
• Computational Chemistry
• Toxicology

Background:

• The soil adsorption coefficient (Koc) is a key environmental fate parameter for risk assessment.
• Current methods for determining Koc are time-consuming and expensive.
• Efficient estimation of Koc is needed early in chemical development.

Purpose of the Study:

• To develop a highly accurate quantitative structure-property relationship (QSPR) model for predicting the soil adsorption coefficient (Koc).
• To utilize calculated physicochemical properties and molecular descriptors for model development.
• To compare the performance of a LightGBM-based model with existing methods.

Main Methods:

• Developed a QSPR model using the largest available Koc dataset.
• Employed the OPEn structure-activity/property Relationship App (OPERA) for physicochemical properties.
• Utilized Mordred software for molecular descriptors.
• Implemented the Light Gradient Boosted Machine (LightGBM) algorithm, a gradient boosting decision tree (GBDT).
• Tuned the LightGBM program for optimal performance.

Main Results:

• The developed QSPR model demonstrated high accuracy in predicting Koc values.
• The combination of LightGBM, OPERA, and Mordred significantly improved Koc prediction accuracy compared to previous models.
• The model effectively analyzes a diverse range of chemical compounds.
• The study reports a successful method for tuning the LightGBM program.

Conclusions:

• A highly accurate QSPR model for Koc prediction was successfully developed using molecular descriptors and physicochemical properties.
• The novel approach using LightGBM, OPERA, and Mordred offers efficient and cost-effective preliminary environmental risk assessment.
• This method is valuable for early-stage chemical development, reducing time and expenditure.