Galux, a South Korean biotech pioneering AI-driven protein therapeutics design, has announced new experimental findings demonstrating that its AI protein design platform, GaluxDesign, can reliably generate high-affinity antibodies from minimal design sets, which the company said is a shift in how therapeutic antibodies can be discovered.
In the new study, the company generated 50 AI-designed antibodies per epitope across eight target epitopes and achieved a 31.5% overall hit rate. Notably, 10.5% of all designs exhibited therapeutically meaningful affinity, with several candidates reaching picomolar binding strength. All candidates were validated in full-length IgG format, confirming that GaluxDesign outputs behave as druggable therapeutic leads without extensive engineering.
“This performance represents more than a numerical improvement in success rate, it reflects a fundamental shift from discovering antibodies to rationally designing them,” said Chaok Seok, CEO of Galux.
“Traditional antibody discovery depends on probabilistic enrichment from massive libraries and requires months of iterative affinity maturation and humanization. The ability of GaluxDesign to deliver potent binders within weeks, and without heavy downstream engineering, demonstrates a true transition from stochastic discovery to rational molecular design.”
The company’s earlier library-scale work demonstrated de novo antibody design across eight distinct therapeutic targets (PD-L1, HER2, EGFR S468R, ACVR2A/B, FZD7, ALK7, CD98hc, IL-11), yielding binders with affinities as strong as 9pM. In that study, Galux resolved a designed PD-L1 antibody complex by cryo-EM with 1.1 Å interface RMSD, validated structural novelty, and showed that functional behaviours such as mutant and subtype specificity could be intentionally encoded and retained in IgG format. Together, the findings show not only that GaluxDesign can produce structurally accurate and functionally tunable antibodies, but also that it can now do so reliably from only dozens of computational sequences.
By unifying structural accuracy, functional developability, target diversity, and now small-design efficiency, GaluxDesign said it provides a foundation for therapeutic discovery that is predictable rather than probabilistic.
“What these studies collectively show is that the essential qualities of a therapeutic antibody can all be defined at the design stage and realized experimentally through our platform,” Seok said.
“Our next step is to take this level of design control toward more advanced modalities, including multi-target binders and new functional architectures, and demonstrate how AI can expand the boundaries of what therapeutic antibodies can do.”
Jim Cornall is editor of Deeptech Digest and publisher at Ayr Coastal Media. He is an award-winning writer, editor, photographer, broadcaster, designer and author. Contact Jim here.