Zero-shot generative models design HER2 antibodies, with 71 low-nanomolar binders among 421 characterised
Source check: Needs source check. Not independent scientific confirmation.
Finding: Shanehsazzadeh et al. (Absci) described generative models for de novo antibody design against HER2 without further optimisation, combined with high-throughput screening of about 10^6 HCDR variants; binding rates were 10.6 percent (HCDR3) and 1.8 percent (HCDR123), 71 of 421 characterised binders had low-nanomolar affinity by SPR, 11 were functionally equivalent or better than trastuzumab, and one showed about 3x higher cell-based potency. Preprint, not peer-reviewed at this version; a later journal version should be checked. Abstract-level review.
Technology: Generative deep learning models for antibody design
Publication: Preprint · Test context: direct test. Catalog categories, not an evidence grade.
- AI contribution
- Generative models proposed HCDR variants that were then screened for binding.
- Test and result
- Designed antibodies were expressed and measured by high-throughput screening and SPR, with comparison to trastuzumab (per abstract).
- Limitation / still unknown
- Peer review, in vivo activity, the reach beyond one target and the model details were not assessed; one organisation authored and tested the designs.
Evidence sources and annotation context
Evidence context (descriptions follow the cited sources; test type is a catalog classification):
Evidence sources: https://europepmc.org/article/PPR/PPR589043
Needs source check · editorial details
The source has not been fully checked or contains conflicting information. Independent confirmation is not assessed by this label.
Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.
Review date: 17 Nov 2026. Review scope: abstract and bibliographic metadata (Europe PMC); full text not assessed.
No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.
Editorial assessment: Impact 4/5 · Validation 3/5 · Novelty 4/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.