An adaptive autoregressive diffusion approach to design active humanized antibodies and nanobodies
8.5
来源：
Nature
关键字：
ML brain science
发布时间：
2025-10-01 23:41
摘要：
HuDiff is an innovative adaptive diffusion approach designed for humanizing antibodies and nanobodies, showing significant advancements in humanness and binding affinity. This method eliminates the need for humanized templates and has been validated through experiments targeting SARS-CoV-2, demonstrating its potential for clinical applications. The research is backed by teams from Lanzhou University and Tencent AI Lab, highlighting its relevance in the biotechnology sector.
原文：
查看原文
价值分投票
评分标准
新闻价值分采用0-10分制，综合考虑新闻的真实性、重要性、时效性、影响力等多个维度。
评分越高，表示该新闻的价值越大，越值得关注。
价值维度分析
domain_focus
1.0
business_impact
0.5
scientific_rigor
1.5
timeliness_innovation
1.5
investment_perspective
2.5
market_value_relevance
1.0
team_institution_background
0.5
technical_barrier_competition
0.5
关键证据
HuDiff generates humanized antibodies that more closely resemble experimentally humanized sequences than existing models.
The best-performing humanized antibody retains binding affinity comparable to the parental antibody.
Neutralization assays confirm that the humanized sequences effectively neutralize the virus.
真实性检查
否
AI评分总结
HuDiff is an innovative adaptive diffusion approach designed for humanizing antibodies and nanobodies, showing significant advancements in humanness and binding affinity. This method eliminates the need for humanized templates and has been validated through experiments targeting SARS-CoV-2, demonstrating its potential for clinical applications. The research is backed by teams from Lanzhou University and Tencent AI Lab, highlighting its relevance in the biotechnology sector.