Zebrafish AC₅₀ modelling: (Q)SAR models to predict developmental toxicity in zebrafish embryo

Area of Science:

• Environmental Toxicology
• Computational Chemistry
• Developmental Biology

Background:

• Developmental toxicity assesses adverse effects on developing organisms from chemical exposure.
• Zebrafish embryo assays are emerging as alternatives to animal testing for teratogenicity.
• In silico modeling of zebrafish developmental toxicity is gaining traction due to data availability.

Purpose of the Study:

• To develop quantitative structure-activity relationship (QSAR) models for predicting developmental toxicity.
• To utilize zebrafish embryo data for in silico chemical safety assessment.

Main Methods:

• Employed a well-defined dataset from the ToxCast Phase I chemical library.
• Developed categorical and continuous QSAR models using gradient boosting machine learning and Monte Carlo techniques.
• Ensured models adhered to Organization for Economic Co-operation and Development principles.

Main Results:

• Classification model achieved 0.89 balanced accuracy and 0.77 Matthews correlation coefficient on the test set.
• Regression model demonstrated 0.70 R-squared in external validation and 0.73 Q-squared in internal validation.
• Statistical quality of both models was satisfactory.

Conclusions:

• Developed robust in silico QSAR models for evaluating developmental toxicity using zebrafish data.
• These models show promise for screening chemical teratogenic potential, supporting reduced animal testing.
• The findings contribute to advancing computational approaches in chemical safety assessment.