E-validation - Unleashing AI for validation.

Abstract

The validation of new approach methods (NAMs) in toxicology faces significant challenges, including the integration of diverse data, selection of appropriate reference chemicals, and lengthy, resource-intensive consensus processes. This article proposes an artificial intelligence (AI)-based approach, termed e-validation, to optimize and accelerate the NAM validation process. E-vali-dation employs advanced machine learning and simulation techniques to systematically design validation studies, select informative reference chemicals, integrate existing data, and provide tailored training. The approach aims to shorten current decade-long validation timelines, using fewer resources while enhancing rigor. Key components include the smart selection of reference chemicals using clustering algorithms, simulation of validation studies, mechanistic validation powered by AI, and AI-enhanced training for NAM education and implementation. A centralized dashboard interface could integrate these components, streamlining workflows and providing real-time decision support. The potential impacts of e-validation are extensive, promising to accel-erate biomedical research, enhance chemical safety assessment, reduce animal testing, and drive regulatory and commercial innovation. While the integration of AI and machine learning offers sig-nificant advantages, challenges related to data quality, complexity of implementation, scalability, and ethical considerations must be addressed. Real-world validation and pilot studies are crucial to demonstrate the practical benefits and feasibility of e-validation. This transformative approach has the potential to revolutionize toxicological science and regulatory practices, ushering in a new era of predictive, personalized, and preventive health sciences.