Merck and Mayo Clinic Partner on AI Drug Discovery
Merck and Mayo Clinic have launched a partnership to accelerate early-stage drug development using AI. The collaboration will combine Merck’s AI and machine learning tools with Mayo Clinic’s extensive clinical datasets and platform. This case study demonstrates the adoption of agentic AI in a highly regulated, data-sensitive industry, requiring robust governance and auditability.
- The collaboration leverages Mayo Clinic's Platform_Orchestrate, a program designed to give partners like Merck direct, secure access to de-identified clinical and multimodal datasets, including biorepositories and registries. This platform approach signifies a shift in healthcare toward enabling external innovation on top of established clinical data infrastructure. - Merck is increasingly employing agentic and generative AI across its operations, with over 80% of its workforce using a proprietary AI platform named GPTeal. This internal platform is used for tasks like accelerating the drafting of clinical study reports from weeks to days, indicating a broad strategy of integrating AI into core workflows beyond just drug discovery. - For governance, AI in pharmaceutical research operates under stringent regulatory frameworks like the FDA's GxP (Good Practice) guidelines, which are being adapted for AI with concepts like Computer Software Assurance (CSA). This requires that AI models are validated for their specific intended use, with a documented audit trail for data, training, and performance, treating the AI as a component of a regulated system. - The use of sensitive patient data necessitates privacy-preserving AI architectures. Techniques like federated learning are being explored to train models across different institutions without centralizing the raw data, addressing HIPAA and GDPR compliance. This approach influences API design, requiring interfaces that can handle decentralized model training and secure aggregation of results. - The partnership is an example of a broader "build on top of" strategy that many CTOs in regulated industries are adopting instead of a simple "build vs. buy" decision. Companies are choosing to leverage established, compliant platforms like Mayo Clinic's and then build their unique, value-adding AI workflows and agentic systems on that foundation, focusing their resources on competitive differentiation rather than foundational infrastructure. - Multi-agent AI systems are a key architectural pattern being explored in pharmaceutical R&D. These systems involve specialized AI "agents" that collaborate on complex tasks such as literature review, molecular design, and trial optimization, mirroring the structure of human research teams and accelerating the preclinical discovery phase. - Emerging regulations, such as the EU AI Act, are classifying some AI applications in drug development as "high-risk," which will mandate rigorous risk assessments, human oversight, and transparency in their design and deployment. This regulatory landscape directly impacts the design of AI platforms and APIs, which must incorporate features for explainability and auditability to ensure compliance. - The collaboration will initially focus on inflammatory bowel disease, atopic dermatitis, and multiple sclerosis. These areas were selected because they can benefit from advanced analytics of multimodal data—including imaging, clinical notes, and genomic data—to develop more tailored therapies.
