CA25152 - LLM4Drugs: Advancing Reliable Evidence for Drug Development with Large Language Models (LLM4Drugs)

Drug development is slow, costly, and prone to late-stage failure – often due to fragmented evidence and missed safety or efficacy signals rather than scientific risk. Integrating data from preclinical studies, clinical trials, regulatory reports, and post-marketing surveillance remains largely manual and too slow for modern decision-making. Large language models (LLMs), able to extract and summarize complex and unstructured biomedical text, could accelerate evidence synthesis but remain unvalidated for regulatory-grade use. A coordinated effort is urgently needed to ensure their safe and effective application.

LLM4Drugs will build a pan-European, inclusive network to make evidence synthesis faster, more reliable, and auditable across the entire drug development pipeline. The Action brings together biomedical and clinical researchers, regulators and health technology assessment (HTA) bodies, the pharmaceutical industry and small and medium-sized enterprises (SMEs), AI developers, ethicists, patients, and non-governmental organizations (NGOs). It will (i) define concrete use cases; (ii) create multilingual benchmark datasets and gold-standard annotations; (iii) develop standard operating procedures (SOPs) and validation protocols aligned with regulatory requirements for traceability, reproducibility, and risk management; and (iv) deliver training, mentorship, and Short-Term Scientific Missions (STSMs) with strong participation from Inclusiveness Target Countries (ITCs) and young researchers and innovators.

The Action complements ongoing European initiatives such as REPO4EU/REPOSYSTEM and evidence synthesis initiatives by focusing specifically on LLM-supported evidence synthesis in drug development. This coordinated effort is timely to prevent fragmented development and establish shared standards that improve transparency, reproducibility, and efficiency; thus, shortening the path from scientific discovery to effective, trusted treatments for patients.