Announcing the Antibody Developability Prediction competition
By John Androsavich, General Manager, Ginkgo Datapoints and Georgia Channing, ML for Science Engineer, Hugging Face
We are happy to announce a first-of-its-kind open benchmarking challenge in ML for antibody engineering: the Antibody Developability Prediction Competition, hosted by Ginkgo Bioworks’ Datapoints team in collaboration with 🤗 Hugging Face.
What is antibody developability and why is it important?
Antibodies have to be manufacturable, stable in high concentrations, and have good safety profiles. Liabilities in these properties can prevent an antibody from reaching the clinic. Collectively, these characteristics are known as developability.
Several computational predictors for antibody developability exist, and many companies rely on internal models. However, there has not yet been an open benchmark where these models can be evaluated on the same held-out set.
What properties are being predicted?
We invite the community to predict developability properties using a subset of assays from the GDPa1 dataset of 246 antibodies:
- 💧 Hydrophobicity
- 🎯 Polyreactivity
- 🧲 Self-association
- 🔥 Thermostability
- 🧪 Titer
There is up to $60k in prizes/credits with separate prizes for each property, and winners will be chosen based on a private held-out set of 80 antibodies at the end of the competition. Anonymous participation is permitted, but winning participants must identify themselves to receive prizes. Submissions are due on November 1st 2025 and winners will be announced shortly afterwards.
Join the community
We aim to build a collaborative community around this challenge. Whether you're competing, sharing insights, or simply staying informed, we welcome your involvement.
How to participate (3 steps)
- Create a Hugging Face account
- Register your team at https://datapoints.ginkgo.bio/ai-competitions/2025-abdev-competition
- Submit your predictions on the competition page
Questions?
📧 Contact us at [email protected] with any questions about the competition — we’ll be monitoring this email closely.