# De novo design

This article presents three efforts from Amazon Bio Discovery: MochiBind, a sequence-only predictor that reframes binding affinity as pairwise comparison aggregated by TrueSkill into a global ranking, achieving higher pairwise accuracy than every structure-based baseline on four held-out antigens and scoring 200,000 antibody pairs in roughly 13 seconds on a CPU; CA-MAP, a context-aware multi-property predictor that uses example antibodies in the prompt to absorb batch offsets, holding a 0.99 correlation under a simulated batch effect where standard fine-tuning falls to 0.

This article presents three efforts from Amazon Bio Discovery: MochiBind, a sequence-only predictor that reframes binding affinity as pairwise comparison aggregated by TrueSkill into a global ranking, achieving higher pairwise accuracy than every structure-based baseline on four held-out antigens and scoring 200,000 antibody pairs in roughly 13 seconds on a CPU; CA-MAP, a context-aware multi-property predictor that uses example antibodies in the prompt to absorb batch offsets, holding a 0.99 correlation under a simulated batch effect where standard fine-tuning falls to 0.

The work introduces LDDM (Large Drug Discovery Model), a unified 3D generative framework supporting constrained and unconstrained docking, fragment linking and growing, and de novo design, together with a programmable design algorithm that produces synthetically accessible compounds satisfying fine-grained objectives; the authors experimentally validated designed or optimised ligands for five therapeutically relevant protein targets, achieving high success rates and identifying molecules with confirmed binding affinity while synthesizing only a small number of generated compounds, with the best designs structurally characterised by NMR spectroscopy and X-ray crystallography indicating high prediction accuracy.

The work introduces LDDM (Large Drug Discovery Model), a unified 3D generative framework supporting constrained and unconstrained docking, fragment linking and growing, and de novo design, together with a programmable design algorithm that produces synthetically accessible compounds satisfying fine-grained objectives; the authors experimentally validated designed or optimised ligands for five therapeutically relevant protein targets, achieving high success rates and identifying molecules with confirmed binding affinity while synthesizing only a small number of generated compounds, with the best designs structurally characterised by NMR spectroscopy and X-ray crystallography indicating high prediction accuracy.