The Antibody Developability Consortium has launched to create a standardized dataset for predicting antibody manufacturability. The initiative aims to address limitations in current predictive models, which often suffer from fragmented and limited data. By pooling resources, the consortium seeks to reduce development time and investment risks for pharmaceutical companies.
Founding members include AbbVie from the United States, argenx from the Netherlands, Lundbeck from Denmark, and Takeda from Japan. The group remains open to new participants. Ginkgo Datapoints manages the scientific execution, including sequence selection and high-throughput characterization, while Apheris provides the federated infrastructure for secure data processing.
The project will compile a dataset of 10,000 antibodies, combining proprietary sequences from members with publicly available data. Apheris allows members to train and benchmark AI models without exposing their raw proprietary data to other participants. Members retain ownership of their contributed data while gaining access to the broader consortium-wide models.
Scientific oversight for the consortium is provided by Professor Charlotte Deane from the University of Oxford and Professor Peter Tessier from the University of Michigan. The group expects to deliver the initial dataset to its members by early 2027. Future plans include expanding the scope to cover more complex antibody formats and additional drug properties.
