G.AI.A: An Integrated Machine-Learning Platform for Predicting Bioaccumulation and Ecotoxicity of Pharmaceuticals.

Area of Science:

• Environmental Chemistry
• Computational Toxicology
• Green Chemistry

Background:

• Pharmaceuticals in aquatic environments pose ecological risks due to bioactive compound accumulation.
• Existing risk assessment methods can be time-consuming and resource-intensive.
• There is a need for predictive tools to support sustainable chemical management and drug design.

Purpose of the Study:

• To develop a computational framework for predicting fish bioconcentration and ecotoxicity.
• To integrate metabolite prediction for comprehensive environmental risk assessment.
• To create an accessible web platform (G.AI.A) for rapid environmental risk evaluation.

Main Methods:

• Development of machine learning models for bioconcentration (BCF) and ecotoxicity (LC50) prediction.
• Incorporation of metabolite prediction using the SyGMa tool.
• Utilized molecular fingerprint analysis for model interpretability.
• Benchmarking of multiple tools for metabolite prediction.

Main Results:

• High-performing ML models achieved ROC AUC scores of 94.60% for bioconcentration and 96.06% for ecotoxicity.
• The framework successfully predicts risks for both parent compounds and their metabolites.
• The G.AI.A web platform provides rapid predictions via SMILES input and supports batch processing.

Conclusions:

• The developed framework and G.AI.A platform advance predictive toxicology for pharmaceuticals.
• This approach supports green-by-design pharmaceutical development and sustainable chemical management.
• The tool enables early-stage environmental risk assessment for regulatory bodies and researchers.