Ginkgo Datapoints and Apheris jointly host a new industry consortium
This consortium is designed to address one of the biggest bottlenecks in biologics development: reliably predicting antibody developability early.
Why this consortium exists
Modern antibody R&D suffers from a fundamental data problem:
Datasets are siloed, non-standardized, and historically generated
Most models fail to generalize beyond their training domain
10,000 + sequences may be needed to train high-performing developability AI models
This consortium solves these challenges by combining large-scale, purpose-built developability data with federated computing.
We are creating collaborative frameworks and helping to establish the standards that will shape the future of AI for the most important areas of drug development
What the consortium offers
A large-scale antibody developability dataset purpose built for ML training
Ginkgo Datapoints generates a massive, harmonized dataset that includes sequences rationally designed to maximize diversity and model performance.
Federated model training across all member datasets
Apheris enables secure, cross-company collaboration, keeping data with their owners and IP protected.
Access to consortium models for internal R&D
Members can use the collaboratively trained models internally for discovery, optimization, and downstream developability assessments
A scalable cross-industry framework
The network structure reflects a growing industry movement toward secure decentralized AI, building on successful precedents such as the AI Structural Biology Network (also hosted by Apheris)
