# De novo design of monoclonal and bispecific antibodies with OFAntibody

Preprint · Valhalla Team

- Archive id: arxiv:2610.09548
- Published: 2026-10-07
- arXiv: https://arxiv.org/abs/2610.09548
- Citations: 0
- Source URL: https://arxiv.org/abs/2610.09548
- PDF: https://arxiv.org/pdf/2610.09548
- Provenance: arxiv

## Abstract

Recent advances in generative protein design have enabled de novo antibody generation with explicit target and epitope conditioning. However, most existing approaches remain formulated around a single antigen-antibody interface, whereas bispecific antibody design requires modeling multi-component complexes in which multiple target-recognition interfaces must coexist and interact within a shared antibody structure. Here we present OFAntibody, an all-atom generative framework for de novo design of monoclonal and bispecific antibodies. OFAntibody expands CDR-epitope interaction learning with large-scale distilled antigen-antibody complexes, and introduces multi-component structural supervision and arm-aware multi-hotspot routing to learn compatible multi-interface geometries and couple each antibody paratope to its designated epitope. OFAntibody supports epitope-conditioned generation across monoclonal antibodies and diverse bispecific formats, including tandem VHH, diabody and CODV. In nanobody design benchmarks, OFAntibody achieves a Top-5 enrichment rate of 41.5%, representing a 5.39-fold improvement over RFantibody in competitive candidate ranking. In bispecific antibody design tasks, OFAntibody achieves hotspot pass rates of 94-100% across the three evaluated tasks and achieves energy pass rates of 60%, 13% and 8% for diabody, tandem VHH and CODV formats, respectively. OFAntibody further enables the same target combination to be explored across different antibody formats, while joint multi-interface generation reduces geometric incompatibilities arising from independent design and post hoc assembly. Together, these results extend de novo antibody design from binary antigen-antibody complexes to programmable multi-component complexes, providing a foundation for designing single molecules that combine recognition of distinct targets and their associated biological functions.

_Abstract reproduced under CC0-1.0 (arXiv descriptive metadata)._

Topics: [AI Protein Design & Structure Prediction](https://space-kitty.com/topics/protein-design)

---
Space Kitty Archive · https://space-kitty.com · API: https://space-kitty.com/api/v1 · MCP: https://space-kitty.com/mcp · Premium: $0.20/request via x402 (Base Sepolia (testnet)). Metadata: arXiv (CC0), OpenAlex (CC0), Europe PMC (CC BY/CC0 works), ClinicalTrials.gov.