Harbour BioMed’s new AI approach to elevate the biologics exploratory study

Image Credit: Harbour Biomed

Announcement

Harbour BioMed, a leading biopharmaceutical company aiming at the establishment and discovery of novel antibody therapeutics in oncology and immunology, has introduced the first and best-in-class fully fledged human generative AI HCAb (Heavy chain-only antibody) model sponsored by its Hu-mAtrlx™ AI platform. This platform stands strong on the harbour mice® platform. This excellent platform develops a relatively closed-loop process unifying smart screening, wet-lab and AI-design approval. This introduction of the new model was announced at the Harbour’s global R&D day 2025 event held in Shanghai.

At the event, the pharmaceutical ecosystem collaboration and global AI confirmed its presence in the near future officially. The partnership gathers comprehensive technologies, resources from numerous sectors, and known expertise under one roof, focusing on sophisticatedly reframe the whole drug R&D process with the help of artificial intelligence technologies. This will elevate the advancement of innovative therapies for patients' well-being and health globally.

An AI approach to sharpen the next-generation biologics discovery

The AI HCAb model develops a closed-loop workflow that amalgamates AI-fuelled sequence generation, we-lab, screening and design validation. This process covers end-to-end operation transitions, antibody exploration from blind screening to AI-powered smart selection. This has eventually enhanced accuracy and efficiency in antibody discovery. The HCAb is popular for its seamless structure and reduced molecular weight that defines its robust capability for advanced applications such as ADCs, mRNA, CAR-T, and bispecific/multi-specific antibody therapies. Though the conventional human VH domains face issues to stand by balancing in the absence of VL pairing.

Harbour Biomed’s Harbour mice platform supporting this new model is the first transgenic mouse platform globally that makes fully human functional HCAb. The major AI design is said to be perfect with its accuracy in addressing autonomous and functional HCAbs from the broader datasets. The recent model elaborates a tenfold benefit in a prominent improved success rate and candidate generation. According to the data from 107 de novo created nbinder sequences carried by the AI HCAb model, it achieved 78.5% of the target, while 20 molecules received a green signal in wet-lab assays.

This emphasises high yield, specificity, activity and purity. Following the AI-developed binders showed spectacular, developable and promising profiles with a yield expanding by 700 mg/L. Various candidate sequences highlighted nanomolar-stage binding affinity, simultaneously retaining binding activity against similar or suitable targets in cynomolgus monkeys and humans. The model aims for the long-term development and improvement in the biologics discovery.