PS26: Breaking Silos: Federated Learning for Next-Gen Drug Development

Collaborative clinical research across sponsors and institutions is often limited by regulatory constraints, data privacy requirements, and proprietary interests. These challenges impede the integration of diverse datasets, reducing the statistical power and generalizability of trial outcomes. Federated learning, a term introduced by Google in 2016, enables decentralized model training across multiple clinical trial sites without transferring sensitive patient-level data. This privacy-preserving approach aligns with frameworks such as HIPAA and GDPR, while fostering collaboration across organizational and disciplinary boundaries. By unlocking access to broader, more representative datasets, federated learning supports improved study design, inclusive patient representation, and more robust outcome prediction. Real-world applications are beginning to demonstrate the promise of federated learning in clinical development. Eli Lilly's TuneLab initiative allows biotech partners to leverage proprietary AI models for drug discovery by exchanging insights through a secure federated framework. The MELLODDY Consortium-uniting ten pharmaceutical companies-has successfully applied federated learning to train predictive models on proprietary compound libraries, enhancing candidate selection while maintaining data confidentiality. In oncology, Owkin has enabled international cancer centers to collaboratively develop treatment response models using real-world patient data, all while ensuring that information remains securely on-site. These efforts exemplify how federated learning can bridge disciplines-uniting clinicians, statisticians, clinical operations, and regulatory experts-to address complex challenges in drug development. This session will feature speakers from regulatory agencies, industry, and academia who will share case studies and methodological advances in federated learning. Topics will include: - Which areas of clinical development benefit most from federated learning? - What are the operational, legal, and resource-related challenges to broader adoption? - How can technical experts collaborate effectively with other functions to advance innovative methodologies? This session is designed for attendees interested in practical implementation and recent advances in federated learning within the clinical development landscape.