Can Artificial intelligence meaningfully shorten drug discovery timelines? Current evidence, roadblocks and future directions
TL;DR
A critical review is needed to distinguish the stages in which AI can improve search, prioritization, and decision-making from those that remain limited by experimental validation, safety assessment, and clinical evidence generation.
Abstract
ABSTRACT Introduction Artificial intelligence (AI) is increasingly proposed as a means of shortening drug-discovery timelines, although its practical impact varies across the discovery and development process. A critical review is needed to distinguish the stages in which AI can improve search, prioritization, and decision-making from those that remain limited by experimental validation, safety assessment, and clinical evidence generation. Areas covered This review discusses the use of AI in target and pathway prioritization, variant and protein-structure interpretation, drug repurposing, virtual screening, molecular design, ADMET prediction, biomarker discovery, and patient stratification. The review was informed by iterative searches of PubMed and Google Scholar, supplemented by ResearchGate, covering literature from database inception to 15 July 2026, together with reference lists, clinical-trial registries, regulatory disclosures, company reports, and publicly available pipeline updates. Expert opinion AI is most likely to shorten drug discovery when it improves the quality and sequence of decisions, reduces the number of unnecessary experiments, and identifies failure earlier. Its greatest value will arise when it is integrated with high-quality data, disease-relevant experimental models, expert supervision, and prospective clinical validation.