Combining Deep Learning and Explainable AI for Toxicity Prediction of Chemical Compounds

2025-10-26

AI Summary

This study addresses compound toxicity prediction in computational toxicology. We propose an interpretable deep learning framework that converts molecular structures into standardized 2D graphical images and feeds them into a pretrained DenseNet121 network for end-to-end toxicity activity prediction. Interpretability is achieved via Grad-CAM, generating molecule-level saliency heatmaps that highlight chemically relevant substructural regions. To our knowledge, this is the first systematic evaluation of image-based representation combined with densely connected convolutional networks for multi-task toxicity prediction on the Tox21 benchmark. The method achieves competitive predictive performance—mean AUC scores of 0.78–0.85—comparable to state-of-the-art machine learning models, while substantially enhancing model transparency and chemical interpretability. By bridging deep learning with domain-specific structural reasoning, our approach establishes a novel paradigm for AI-driven mechanistic toxicity investigation.

Technology Category

• Computer Vision: Interpretability, Explainability, and Transparency
• Machine Learning: Transparent, Interpretable, Explainable ML
• Knowledge Representation and Reasoning: Computational Complexity of Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

The task here is to predict the toxicological activity of chemical compounds based on the Tox21 dataset, a benchmark in computational toxicology. After a domain-specific overview of chemical toxicity, we discuss current computational strategies, focusing on machine learning and deep learning. Several architectures are compared in terms of performance, robustness, and interpretability. This research introduces a novel image-based pipeline based on DenseNet121, which processes 2D graphical representations of chemical structures. Additionally, we employ Grad-CAM visualizations, an explainable AI technique, to interpret the model's predictions and highlight molecular regions contributing to toxicity classification. The proposed architecture achieves competitive results compared to traditional models, demonstrating the potential of deep convolutional networks in cheminformatics. Our findings emphasize the value of combining image-based representations with explainable AI methods to improve both predictive accuracy and model transparency in toxicology.

Problem

• Predicting chemical compound toxicity using deep learning
• Interpreting toxicity predictions with explainable AI techniques
• Improving accuracy and transparency in computational toxicology models

Innovation

• DenseNet121 processes chemical structure images
• Grad-CAM visualizations interpret toxicity predictions
• Image-based pipeline with explainable AI methods