Apheris, a Berlin-based artificial intelligence company, has announced a significant collaboration with Ginkgo Datapoints to establish the Antibody Developability Consortium, a joint initiative designed to accelerate antibody drug development through machine learning and standardized data sharing.
The consortium brings together a coalition of major pharmaceutical companies, including AbbVie, argenx, Lundbeck, and Takeda, who will contribute proprietary antibody sequences to build what organizers describe as the industry’s largest standardized dataset of 10,000 antibodies. This pooled resource will enable the development of artificial intelligence models specifically trained to predict manufacturability challenges and other developability risks associated with antibody therapeutics.
Secure Data Sharing at Scale
A central innovation of the initiative lies in its approach to data collaboration. Rather than requiring consortium members to share raw antibody sequences—sensitive intellectual property in the pharmaceutical industry—Apheris will leverage its proprietary federated learning infrastructure. This technology allows participating organizations to train and fine-tune AI models using their own data without exposing underlying sequences to other members or external parties.
The federated approach addresses a persistent challenge in pharmaceutical research: the tension between the competitive nature of drug development and the need for large, diverse datasets to train effective machine learning algorithms. By maintaining data privacy while enabling collaborative model development, the consortium creates an environment where competing firms can jointly advance scientific progress without compromising confidential information.
Accelerating Early-Stage Development
The consortium’s primary objective centers on improving the early prediction of developability risks in antibody programs. Identifying potential manufacturing or stability issues at preliminary stages of drug development can significantly reduce timelines and costs associated with later-stage failures. AI models trained on a comprehensive dataset of diverse antibodies could theoretically flag problematic characteristics before substantial resources are invested in clinical programs.
Robin Röhm, commenting on the initiative’s significance, noted that practical implementation remains crucial. “For AI to impact developability decisions in a drug program, it has to perform on a pharma’s own molecules,” Röhm stated, underscoring that general models must demonstrate effectiveness when applied to individual company data and workflows.
Broader Ecosystem Implications
The Antibody Developability Consortium reflects growing momentum in European biotech toward collaborative AI applications. Germany has emerged as a regional hub for AI-driven drug discovery, with Berlin particularly attracting startups focused on computational biology and pharmaceutical innovation. Similar consortia models—combining academic institutions, startups, and established pharmaceutical companies—have gained traction across Europe as organizations recognize that pooled expertise and data can accelerate breakthrough discoveries while maintaining competitive advantages.
The initiative demonstrates how federated learning technologies can facilitate knowledge-sharing at scale, potentially serving as a template for other therapeutic areas beyond antibody development where companies face similar data-sharing constraints.