Summary
Antibody-antigen complex prediction appears to scale with training data and compute. Co-folding performance appears to follow a scaling law dependent on both training data and compute (1).
Figures
Ref (1)
See also
- Antibody-antigen modeling by diffusion-based structure prediction is data-limited
- Scaling hypothesis
- Correct antibody-antigen prediction in AF3 and related models is partially determined by training set similarity
1.
Jing B, Bafna M, Diaz D, Klivans A, Berger B. Promera: a unified model for biomolecular structure prediction, filtering, and design. openRxiv; 2026. Available from: https://doi.org/10.64898/2026.06.07.729267