AI in Drug Discovery: Clinical Failures, Regulatory Reality, and the Validation Crisis Behind the Hype
- 1Department of Diagnostic and Allied Health Science, Faculty of Health and Life Sciences, Management and Science University, Shah Alam 40100, Selangor, Malaysia
- 2Neuroscience and Mental Well-Being Centre (NeuroMIND), Management and Science University, Shah Alam 40100, Selangor, Malaysia
Abstract
The integration of artificial intelligence (AI) into the life sciences has accelerated significantly between 2022 and 2026, accompanied by global investment exceeding USD 100 billion and widespread expectations of a transformative impact in drug discovery. Despite these advances, the extent to which AI has improved clinical outcomes remains unclear. This study presents a structured narrative review evaluating the economic, technical, clinical, and regulatory dimensions of AI adoption in drug discovery. Current evidence indicates that clinical attrition rates remain high, with approximately 90% of drug candidates entering clinical development failing to achieve regulatory approval. Although AI systems such as AlphaFold have achieved high structural prediction accuracy, with predicted local distance difference test (pLDDT) scores exceeding 90 for well-structured proteins and root mean square deviation (RMSD) values comparable to experimental methods, limitations persist in modelling protein dynamics, post-translational modifications, and protein-ligand interactions. Clinical case studies demonstrate that while AI can accelerate early-stage discovery timelines, these advantages do not consistently translate into improved late-stage success rates. Furthermore, reproducibility challenges, limited data transparency, and regulatory gaps continue to constrain reliable implementation. These findings suggest that AI in drug discovery is currently in a transitional phase characterised by high investment but limited validated clinical impact. Future progress will depend on strengthening validation frameworks, improving data sharing practices, and aligning regulatory standards with real-world clinical performance.
Related
- Integrating artificial intelligence and molecular simulation for peptide discovery: limitations, validation and decision-centered workflows
- PolyDiscover: an integrated trust-aware platform for multi-objective polymer discovery and simulation validation
- Integrating machine learning-based molecular design with experimental validation for the discovery of EGFR inhibitors in lung cancer
- A resource-efficient structure-based workflow for fragment progression enables parallel hit discovery and validation of functionally diverse modulators of NCS-1 protein–protein interactions
- Computationally Guided Discovery of Cannabinoid Derivatives for Antiviral and Anticancer Applications: From Chemical Space to Experimental Validation
BibTeX
@article{khairil2026drug,
title = {{AI in Drug Discovery: Clinical Failures, Regulatory Reality, and the Validation Crisis Behind the Hype}},
author = {Lisa Khairil and Koay Hean Seng Benny and Jesreena Jerry and Farhat Mussa Khatib and Muhammad Danial Che Ramli and Suresh Kumar},
journal = {Pharmaceuticals},
year = {2026},
month = jun,
doi = {10.3390/ph19060916},
url = {https://doi.org/10.3390/ph19060916},
note = {CC BY},
}
RIS
TY - JOUR
TI - AI in Drug Discovery: Clinical Failures, Regulatory Reality, and the Validation Crisis Behind the Hype
AU - Lisa Khairil
AU - Koay Hean Seng Benny
AU - Jesreena Jerry
AU - Farhat Mussa Khatib
AU - Muhammad Danial Che Ramli
AU - Suresh Kumar
PY - 2026
DA - 2026/06/10
JO - Pharmaceuticals
DO - 10.3390/ph19060916
UR - https://doi.org/10.3390/ph19060916
AB - The integration of artificial intelligence (AI) into the life sciences has accelerated significantly between 2022 and 2026, accompanied by global investment exceeding USD 100 billion and widespread expectations of a transformative impact in drug discovery. Despite these advances, the extent to which AI has improved clinical outcomes remains unclear. This study presents a structured narrative review evaluating the economic, technical, clinical, and regulatory dimensions of AI adoption in drug discovery. Current evidence indicates that clinical attrition rates remain high, with approximately 90% of drug candidates entering clinical development failing to achieve regulatory approval. Although AI systems such as AlphaFold have achieved high structural prediction accuracy, with predicted local distance difference test (pLDDT) scores exceeding 90 for well-structured proteins and root mean square deviation (RMSD) values comparable to experimental methods, limitations persist in modelling protein dynamics, post-translational modifications, and protein-ligand interactions. Clinical case studies demonstrate that while AI can accelerate early-stage discovery timelines, these advantages do not consistently translate into improved late-stage success rates. Furthermore, reproducibility challenges, limited data transparency, and regulatory gaps continue to constrain reliable implementation. These findings suggest that AI in drug discovery is currently in a transitional phase characterised by high investment but limited validated clinical impact. Future progress will depend on strengthening validation frameworks, improving data sharing practices, and aligning regulatory standards with real-world clinical performance.
SN - 1424-8247
ER -