跳到正文
Molence· @MolenceBio · X·· 17 天前AI 评分56
AI 导读

Molence 的 Mol-1-constraint 把表位信息作为预测输入,在四个抗体-抗原基准上均提升了 DockQ > 0.23 的成功率。FoldBench-AB 从 75.0% 升至 88.3%,AF3-AB 从 62.7% 升至 84.2%,PXMeter-AB 从 58.3% 升至 77.5%,MolBench-AB-26 从 57.3% 升至 79.4%。基础模型和 Mol-1-constraint 均可在 https://lab.molence.com 获取。

正文 · 原文

14/15 📍 An epitope can be an input to prediction.

With epitope information, Mol-1-constraint improves antibody–antigen success at DockQ > 0.23 across all four benchmarks:
• FoldBench-AB: 75.0% → 88.3%
• AF3-AB: 62.7% → 84.2%
• PXMeter-AB: 58.3% → 77.5%
• MolBench-AB-26: 57.3% → 79.4%

Both the base model and Mol-1-constraint are available at https://lab.molence.com

#AIforScience

引用Molence@MolenceBio
13/15 🧬 Protein–peptide prediction, through the high-accuracy threshold. On MolBench-Peptide-26, Mol-1 achieves: • 88.8% at DockQ > 0.23 • 77.3% at DockQ > 0.49 • 54.1% at DockQ > 0.80 With binding-site information, Mol-1-constraint raises these to 92.2%, 85.0% and 60.2%. Biological site information improves peptide-complex prediction at every evaluated accuracy threshold. #AIforScience
在 X 查看被引用的帖子

来源:Molence · x.com