Artificial intelligence and machine learning in drug discovery and toxicology

Dragos Cretoiu ^{1, 2 *}

¹ Department of Medical Genetics, Carol Davila University of Medicine and Pharmacy, Bucharest, Romania
² Materno-Fetal Medicine Excellence Center, Alessandrescu-Rusescu National Institute of Mother and Child Health, Bucharest, Romania
* Correspondence to: Dragos Cretoiu, Materno-Fetal Medicine Excellence Center, Alessandrescu-Rusescu National Institute of Mother and Child Health, Bucharest, Romania. E-mail: dragos@cretoiu.ro

Abstract

Traditional drug discovery is slow, costly, and prone to high attrition, with only a fraction of candidate compounds surviving the arduous journey from bench to bedside. Artificial intelligence (AI) and machine learning (ML) are now being positioned as transformative forces in this domain, capable of accelerating target identification, optimizing lead molecules, and predicting safety profiles earlier in development. This editorial explores the current state of AI in drug discovery and toxicology, drawing on concrete examples such as AI-driven drug repurposing during COVID-19, the rapid progression of generative AI-designed molecules into clinical trials, and the emergence of computational toxicology as a credible alternative to animal testing. We highlight the regulatory and ethical landscapes, noting both progress—such as the FDA’s 2025 draft guidance on AI models—and persistent challenges, including model interpretability, bias, and dual-use concerns. AI and ML will not replace human drug hunters, but they are reshaping the contours of pharmaceutical research and development, offering a vision of faster, more humane, and potentially more cost-effective innovation. Realizing this vision will require responsible integration: validating models rigorously, embedding transparency and equity, and ensuring that human oversight remains central.