Bristol Myers Squibb and Takeda Pharmaceuticals have agreed to collaborate in a new AI-driven drug discovery initiative, pooling proprietary molecular structure datasets to train an advanced AI model known as OpenFold3. Under the arrangement, both companies will contribute thousands of experimentally determined protein–small molecule structures via a federated, privacy-preserving platform managed by Apheris.

This partnership forms part of the broader AI Structural Biology (AISB) Network, which also includes firms such as AbbVie and Johnson & Johnson. The federated model enables participating companies to improve predictive capabilities—especially with respect to protein–ligand interactions—without disclosing proprietary data. By sharing diverse datasets securely, Bristol Myers and Takeda aim to accelerate the pace of early discovery and reduce the duplication of effort across the industry.

Payal Sheth of Bristol Myers noted that collaborative AI models surpass what individual organisations can achieve alone, while Hans Bitter from Takeda emphasised the alignment with each company’s strategic goal of embedding AI across their discovery pipelines. The partnership reflects a growing trend in pharma, where companies are increasingly relying on advanced computational tools and shared initiatives to tackle the complexity of modern drug discovery.