Harbour BioMed Introduces First Fully Human Generative AI HCAb Model to Advance Biologics Discovery
Wednesday, October 29, 2025
Harbour BioMed a global biopharmaceutical company specialising in antibody therapeutics for immunology and oncology, has launched its first fully human Generative AI HCAb (Heavy Chain-Only Antibody) Model. The model is powered by the company’s Hu-mAtrIx™ AI platform, which is built upon the Harbour Mice® technology. This integrated platform creates a closed-loop system that combines AI design, intelligent screening, and laboratory validation.
The announcement was made during Harbour BioMed’s Global R&D Day 2025 in Shanghai, where the company also introduced the Global AI + Pharmaceutical Ecosystem Alliance. The alliance aims to unite experts, technologies, and resources from various sectors to transform the drug research and development process using artificial intelligence, with the goal of accelerating innovative therapies worldwide.
The newly developed AI HCAb Model provides an end-to-end process that links AI-driven sequence generation, intelligent screening, and laboratory validation. This approach replaces traditional blind screening with precise AI-guided selection, significantly improving the accuracy and speed of antibody discovery.
HCAbs, known for their simple structure and low molecular weight, hold strong potential for advanced applications such as bispecific or multi-specific antibodies, ADCs, CAR-T therapies, and mRNA-based treatments. Conventional human VH domains often face stability challenges without VL pairing, but Harbour BioMed’s Harbour Mice® platform— the first transgenic mouse model that produces fully human functional HCAbs — has addressed this issue.
The AI HCAb model is trained on nine million NGS-derived HCAb sequences, along with extensive public datasets. Using a fine-tuned protein language model, it enables the de novo generation of HCAb sequences with secondary optimisation for target specificity. These sequences undergo a multi-stage AI screening process, which includes classification, developability assessment for stability and solubility, and final laboratory validation.
The AI HCAb Model achieved a tenfold increase in candidate generation and a notable improvement in binding success rates. Among 107 de novo binder sequences generated by the model, 78.5% successfully targeted their intended molecules, and 20 candidates were further validated through laboratory testing. These candidates showed high activity, purity, and binding specificity.
Several AI-designed binders demonstrated nanomolar-level binding affinity with strong performance in both human and cynomolgus monkey targets. The average protein yield exceeded 700 mg/L, indicating excellent developability.
The Hu-mAtrIx™ AI platform operates as a self-learning system, integrating AI design, automated validation, and re-learning. This iterative process enables continuous improvement in model intelligence and R&D productivity.
With this innovation, Harbour BioMed aims to advance the development of fully human HCAbs across multiple therapeutic areas, including multi-specific antibodies, XDCs, in vivo CAR-T, and oral or inhaled biologics.
The newly launched Global AI + Pharmaceutical Ecosystem Alliance will promote collaboration among researchers, technology partners, and investors to modernise the drug discovery pipeline through AI integration.
The alliance has received support from several organisations, including Fortera Capital, Insilico Medicine, Molecular Mind, Evinova, INNOVEL, Fenglin Group, Taimei Technology, EClinCloud, Deep Intelligent Pharma, and Harbour BioMed.
According to Harbour BioMed, the initiative aims to foster an open and collaborative environment that encourages innovation, helping to overcome key challenges in drug discovery and deliver transformative therapies to patients more efficiently.
Source: prnewswire.com
