Modeling bioconcentration factor (BCF) using mechanistically interpretable descriptors computed from open source tool...

Area of Science:

• Environmental Chemistry
• Computational Chemistry
• Toxicology

Background:

• Bioconcentration factor (BCF) is crucial for assessing aquatic chemical toxicity.
• Predictive models are needed to estimate BCF efficiently.
• Mechanistically interpretable descriptors can improve model reliability.

Purpose of the Study:

• To develop and validate predictive regression models for BCF.
• To utilize extended topochemical atom (ETA) indices and XLogP as descriptors.
• To employ open-source software (PaDEL-Descriptor) for descriptor calculation.

Main Methods:

• A dataset of 522 diverse chemicals was used.
• Extended topochemical atom (ETA) indices and XLogP were calculated using PaDEL-Descriptor.
• Genetic function approximation and partial least squares analyses were applied for model development.

Main Results:

• Models incorporating ETA indices and XLogP demonstrated excellent statistical quality and predictive ability.
• Models identified key descriptors influencing BCF, including lipophilicity, heteroatoms, and halogens.
• External dataset predictions yielded high R² values (0.812 and 0.826).

Conclusions:

• Predictive models for BCF can be effectively developed using ETA and XLogP descriptors.
• PaDEL-Descriptor software provides a reliable tool for calculating necessary chemical descriptors.
• These models support aquatic chemical toxicity management by enabling accurate BCF predictions.