Antigen-specific Antibody Multi-modal Foundation Model for Functional Antibody Design
AuthorsXiaoliang Shi, Zichen Wang, Runze Ma, Zhongyue Zhang, Shuangjia Zheng
AAMFM is a multimodal foundation model that uses antigen structure and epitope context to design more functional, antigen-specific antibodies.
Key results
Parameter scale of the multimodal foundation model used as AAMFM’s base.
Paired antibody sequence-structure examples used for antibody-domain adaptation.
Parameters in the GearNet and epitope-aware cross-modal adapter.
Approximately this many preference pairs were generated for preference alignment.
AAMFM-CalDPO’s AF3 score on the SAbDab full-antibody design task.
AAMFM-CalDPO’s predicted interface confidence on full-antibody design.
What the paper found
Researchers at Shanghai Jiao Tong University introduce AAMFM, an antigen-specific antibody multimodal foundation model built on the 1.4B-parameter ESM3-open model. AAMFM jointly represents antibody sequences and structures while conditioning generation on antigen geometry and binary epitope annotations through a GearNet-based cross-modal adapter containing 5.1M parameters. Training proceeds from approximately 1.4M paired antibody sequence-structure examples in OAS to experimentally resolved antibody-antigen complexes from SAbDab, followed by Calibrated Direct Preference Optimization, or Cal-DPO. For preference alignment, the team uses Protenix, ByteDance’s open reproduction of AlphaFold3, to score predicted antibody-antigen complexes, and AntiBERTy pseudo-log-likelihood to constrain sequence plausibility; roughly 30k preference pairs are formed using dual thresholds. On SAbDab full-antibody design, AAMFM-CalDPO reaches an AF3 score of 0.892 and an ipTM of 0.888, leading all compared methods on these functional and structural measures. In the CDR-H3-specific task, it achieves an AF3 score of 0.894 and ipTM of 0.890, also surpassing the baselines. Ablations show that antigen geometric conditioning is especially important, while Cal-DPO improves predicted plausibility, foldability, and binding confidence. The results are computational rather than experimental: the designed antibodies have not yet undergone in-vitro validation.
Original abstract
Antibodies are essential proteins that play a central role in immune recognition by binding specific antigen molecules. Although recent protein language models have enabled progress in single-chain protein modeling and generation, they often fall short in antigen-specific antibody design, where effective modeling requires explicit pairing between antibody and antigen, particularly at the epitope level. To address these limitations, we introduce AAMFM, an Antigen-specific Antibody Multimodal Foundation Model that learns unified representations of antibody sequences and structures conditioned on antigen context. AAMFM incorporates rich antigen information including geometric interfaces and epitope annotations via a cross-modal adapter, enabling joint modeling of antibody-antigen interactions in a shared latent space. To further guide the model toward functional relevance, we fine-tune AAMFM using Calibrated Direct Preference Optimization (Cal-DPO), leveraging preference signals extracted from a strong structural prior to align learning with binding-specific objectives. Extensive experiments demonstrate that AAMFM achieves state-of-the-art performance in functional antibody design, revealing its potential for antigen-specific antibody engineering. Our code is available at https://github.com/XL-S224/AAMFM.
Read the original paperMore in AI for Science
Browse all 43 papers →AI-guided high-throughput discovery of iridium- and ruthenium-free palladium-oxide catalysts for durable acidic oxygen evolution
Ken J. Jenewein, Faezeh Habib Zadeh, Xiaoxiao Wang, Gustavo Malkomes, Huafan Zhang, Natalie Page, Jae Jin Bang, Peter J. Santiago, Karla V. Contreras, Katherine K. Li, Allison Perna, Lorena M. Britton, Fahrettin Kilic, Kevin J. Cruse, Armin Taheri, Krishnanand Mallayya, Harley Quinn, Rebecca A. Durr, Peter A. Beaucage, John M. Gregoire, Rafael Gómez-Bombarelli
An AI-guided robotic lab discovered palladium-based catalysts that could make acidic water electrolysis more durable while reducing dependence on scarce iridium and ruthenium.
Discovery of radio emission from the exoplanet $β$ Pictoris b
Kevin N. Ortiz Ceballos, Edo Berger, Yvette Cendes
Astronomers have detected radio auroras from β Pictoris b, revealing that this distant giant planet has a magnetic field at least 1.25 kilogauss strong.
EurekaBench: Measuring Agentic Ability to Discover New Scientific Insights
Jiayi Geng, Zhengxuan Wu, Kevin S. Chen, Seungone Kim, Joseph Janssen, Zora Zhiruo Wang, Bhupalee Kalita, Runtian Gao, Aaron Ho, Andrew Oakleigh Nelson, Olexandr Isayev, Francisco Villaescusa-Navarro, Ching-Yao Lai, Howard Chen, Graham Neubig
EurekaBench tests whether AI agents can move beyond accurate prediction to uncover mechanisms and insights that genuinely advance scientific understanding.