QSAR Models for Repeated Dose Toxicity in Rats Using the CORAL Software.

Area of Science:

• Toxicology
• Computational Chemistry
• Pharmacology

Background:

• Evaluating chemical safety necessitates determining safe dosage levels, typically via prolonged animal studies.
• The No Observed Adverse Effect Level (NOAEL) represents the highest dose without adverse effects, a key metric in toxicity assessment.
• Existing data for NOAEL determination is extensive, necessitating efficient analysis methods.

Purpose of the Study:

• To develop in silico models for predicting repeated-dose toxicity in rats.
• To accelerate the assessment of toxicity for a large number of chemical substances.
• To establish a computational approach for identifying the No Observed Adverse Effect Level (NOAEL).

Main Methods:

• Utilized experimental NOAEL data from literature and the OpenFoodTox database (n=848).
• Applied a Monte Carlo technique combined with the Las Vegas algorithm for model development.
• Calculated optimal molecular descriptors using correlation weights from Simplified Molecular Input Line Entry System (SMILES) attributes.

Main Results:

• Developed predictive models for repeated-dose toxicity in rats.
• Achieved a good predictive potential with an average determination coefficient of 0.77 ± 0.04 on the validation set.
• Demonstrated the efficacy of in silico methods in toxicity assessment.

Conclusions:

• In silico models offer an attractive and efficient solution for predicting chemical substance toxicity.
• The developed models show significant predictive potential for NOAEL determination.
• Computational approaches can expedite the safety evaluation of chemicals, complementing traditional methods.