Apheris and Ginkgo unite four pharma giants to train AI on 10,000 antibodies
What's the deal? Berlin-based Apheris and Ginkgo Datapoints launched the Antibody Developability Consortium, a collaboration pooling proprietary drug data from four pharmaceutical companies to train AI models. Founding members are AbbVie, argenx, Lundbeck, and TakedaDealroom has a profile for this one. Try Dealroom →. The group aims to build the field's largest standardised antibody developability dataset.
What's the endgame? The consortium wants to predict earlier which antibody drugs can actually be manufactured and advanced to the clinic. Each member contributes proprietary antibody sequences, with Ginkgo filling any gaps from public sources to reach 10,000 antibodies total. It plans to deliver an initial dataset to members by early 2027.
Why now? Developability problems — the biophysical properties that determine whether an antibody can be made and formulated — can stall promising candidates. Existing predictive models have been held back by small, fragmented datasets, while even large internal datasets lack sequence diversity.
How does it work? Using Apheris's federated infrastructure, members can train, benchmark, and refine AI models on the full dataset without exposing raw proprietary sequences to one another. Members retain ownership of the sequences and assay data they contribute, and can apply the resulting models internally.
Ginkgo Datapoints leads the scientific work: designing sequence selection, overseeing antibody production, and running high-throughput wet-lab characterisation. It also trains a foundation antibody developability model within Apheris's secure environment.
Charlotte Deane, professor of structural bioinformatics at the University of Oxford, and Peter Tessier, professor of pharmaceutical sciences and chemical engineering at the University of Michigan, will provide independent scientific oversight.
What the founders say: "For AI to impact developability decisions in a drug program, it has to perform on a pharma's own molecules," said Robin Röhm, chief executive officer and co-founder of Apheris.
The signal: The consortium reflects a growing push to make privacy-preserving, federated data-sharing work in drug discovery, where competitors sit on valuable but siloed datasets. By pooling data without surrendering ownership, rivals can build shared AI capability neither could reach alone — and the group remains open to additional members.
Read more: Tech.eu
