GNNs Fall Short in Predicting Drug Toxicity, Study Finds

A new study published on arXiv explores the limitations of Graph Neural Networks (GNNs) in predicting drug toxicity, specifically focusing on acetylsalicylic acid (Aspirin). The research found that molecular structure alone can only explain about 45% of Aspirin's known adverse effects. To address this, the study introduces a four-category taxonomy to classify explainability gaps, highlighting issues such as non-encodable effects, missing data, assay mismatches, and representation errors. These findings have significant implications for drug safety monitoring and regulatory practices. AI

IMPACT Highlights the limitations of current AI models in fully capturing drug safety signals, suggesting a need for more comprehensive data and methodologies in pharmaceutical research.

RANK_REASON The cluster contains a research paper detailing a new taxonomy for explainability gaps in GNN-based drug toxicity prediction. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• ChEMBL
• drug toxicity prediction
• GNNExplainer
• Good Pharmacovigilance Practice
• Graph Neural Networks
• Tox21 benchmark

• paper
• safety

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →