Amazon ScienceMelih Yilmaz, Luca Giancardo, Yue Zhao, Edward Lee, Chuanyui Teh, Fangda Xu, Gordon Trang10 min readadvanced
Advancing AI for biology: Teaching models to design and characterize antibodies
Summary
Amazon Bio Discovery developed three AI models: MochiBind for fast, sequence-based antibody binding ranking, CA-MAP for context-aware developability prediction robust to batch effects, and an agent-guided system for de novo antibody design. These advancements aim to accelerate and improve the accuracy of antibody drug discovery, with experimental validation for a novel cancer target.
- MochiBind predicts relative antibody binding affinity from sequence, achieving 10% higher accuracy and 100x faster inference than structure-based methods.
- CA-MAP uses context-aware learning to predict antibody developability properties, maintaining 0.99 correlation despite batch effects and running 200x faster than LLMs.
- An agent-guided de novo design system successfully generated 46 validated strong binders against a novel cancer target with no prior structural data.
- The work addresses key bottlenecks in antibody engineering: binding prediction, developability, and end-to-end design.
This work matters to biopharmaceutical engineers and researchers as it offers concrete AI advancements to significantly reduce the cost and time of developing new antibody-based drugs.
7/10
