De novo generation of SARS-CoV-2 antibody CDRH3 with a pre-trained generative large language model
Problem
Discovering antigen-specific antibodies relies on isolating them from serum, a resource-intensive and time-consuming process.
Why it matters
Slow antibody discovery limits rapid response to pathogens such as SARS-CoV-2 and its variants.
What the AI did
PALM-H3, a pre-trained generative large language model, generates de novo antibody CDRH3 sequences; A2binder pairs epitope and antibody sequences to predict binding specificity and affinity.
Results
PALM-H3-generated antibodies bound SARS-CoV-2 antigens including the XBB variant, with high binding affinity confirmed by in-silico analysis and in-vitro assays.
Impact
Demonstrates AI-generated antibodies with validated binding in vitro; clinical or therapeutic impact not stated.
Record
| Venue | Nature Communications |
|---|---|
| Published | 2024 |
| Work type | article |
| Authors | Haohuai He, Bing He, Lei Guan, Yu Zhao, Feng Jiang, Guanxing Chen, Qing-Ge Zhu, Calvin Yu‐Chian Chen et al. |
| Citations | 69 (OpenAlex, at import time) |
| References | 77 |
| Model or method | PALM-H3; A2binder |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41467-024-50903-y |
| Open access | gold |
| Keywords | antibody design, generative language model, CDRH3, SARS-CoV-2, in vitro validation |
Importance rating
significant — thousands to millions affected
Abstract ties slow serum-based antibody discovery to rapid response to pathogens like SARS-CoV-2 variants; generated antibodies validated in vitro, but clinical/therapeutic impact not stated.