Antibody developers want to know early whether a promising molecule can survive manufacturing. A consortium run by Ginkgo Bioworks’ Datapoints unit and Berlin-based Apheris aims to answer that question with pooled data rather than guesswork.
The effort is called the Antibody Developability Consortium. AbbVie, argenx, Lundbeck and Takeda joined as founding members, and the organizers say other pharma and biotech companies can still sign on. Each of the four hands over proprietary antibody sequences, while public sources fill whatever room remains.
Developability is shorthand for the biophysical traits that decide whether a candidate can be produced at scale and hold up long enough to reach patients. The partners describe their project as the field’s largest standardized set of its kind, assembled with machine learning in mind instead of convenience.
The partners are aiming to gather 10,000 antibodies in all, blending member contributions with what the group draws from public repositories.
Apheris contributes the federated infrastructure that lets each member train on the group’s combined data without exposing its own sequences. AbbVie’s Athena Hadjixenofontos said the arrangement answers the shortcomings of datasets collected for ease rather than for machine learning.
Ginkgo Datapoints leads the scientific design, from sequence selection to antibody production and high-throughput lab characterization, and the group has brought in outside scientific advisers from Oxford and the University of Michigan. Membership remains open to additional pharma and biotech firms.
