Bristol Myers Squibb, Takeda Pharmaceuticals, and Astex Pharmaceuticals are joining a major pharmaceutical consortium to advance artificial intelligence-based drug discovery through collaborative data sharing. The companies announced their participation in an initiative that includes AbbVie and Johnson & Johnson, aimed at training an AI model called OpenFold3 to enhance drug development capabilities.
Federated Data Sharing Model Enables Secure Collaboration
The consortium operates through a federated data sharing model that allows pharmaceutical companies to collaborate without exposing sensitive proprietary information. Germany-based life sciences company Apheris facilitates this collaboration through its computing platform, which enables the aggregation of diverse datasets while ensuring each dataset remains securely in its original location.
The participating companies will contribute data from several thousand experimentally determined protein-small molecule structures to train the OpenFold3 AI model. This pooled approach aims to significantly improve the model's accuracy in predicting interactions between proteins and small molecules, a critical component of drug discovery.
Industry Leaders Emphasize Collaborative Benefits
"The federated platform allows multiple companies to advance predictive models for small molecule discovery in ways no single organization could achieve alone," said Payal Sheth, vice president of discovery biotherapeutics and lead discovery and optimization at Bristol Myers Squibb.
Hans Bitter, head of computational sciences at Takeda, emphasized the strategic importance of the collaboration: "This consortium really ties into our larger corporate goal of embedding AI throughout all of what we do; and also a nice example of how we can come together as pharma companies and do even more for patients than we could if we did it on our own."
OpenFold3 as Flagship AI Initiative
OpenFold3 represents the flagship project of the industry-led AI Structural Biology Network, developed in collaboration with the AlQuraishi Lab at Columbia University. The model is designed to enhance the prediction of molecular interactions, potentially accelerating the identification and optimization of new therapeutic compounds.
The consortium's approach addresses a fundamental challenge in pharmaceutical research: the need for large, diverse datasets to train effective AI models while maintaining competitive advantages and protecting proprietary information. By leveraging federated learning principles, the initiative enables companies to benefit from collective intelligence without compromising individual data security.
