CORAL: building up the model for bioconcentration factor and defining it's applicability domain

Area of Science:

• Computational chemistry
• Cheminformatics
• Toxicology

Background:

• Quantitative structure-activity relationships (QSAR) are crucial for predicting chemical properties.
• CORAL software facilitates QSAR model development using simplified molecular input line entry system (SMILES) descriptors.
• Ensuring model reliability and applicability domain is vital for accurate predictions.

Purpose of the Study:

• To evaluate the applicability domain of QSAR models using CORAL software.
• To introduce a novel function for improving QSAR model predictivity and reliability.
• To enhance the performance of a bioconcentration factor (logBCF) QSAR model.

Main Methods:

• Utilized CORAL software for QSAR model development.
• Employed simplified molecular input line entry system (SMILES) for descriptor calculation.
• Introduced a new function based on prediction errors (Delta(obs)) to identify and eliminate outliers from the training set.

Main Results:

• Successfully evaluated the applicability domain of a logBCF QSAR model.
• The new outlier elimination function significantly improved model predictivity.
• The enhanced model demonstrated increased reliability for new chemical compounds.

Conclusions:

• The developed approach effectively enhances QSAR model performance.
• Outlier identification and removal are critical for robust QSAR modeling.
• This method offers a valuable tool for improving the prediction accuracy of chemical properties.