[论文] EvoStruct: Bridging Evolutionary and Structural Priors for Antibody CD...

论文概要

研究领域: ML 作者: Mansoor Ahmed, Sujin Lee, Umar Khayaz 发布时间: 2025-05-20 arXiv: 2505.15985

中文摘要

等变图神经网络(GNN)方法在抗体互补决定区(CDR)设计中实现最高序列恢复率，但遭受严重词汇崩溃。当前最佳GNN方法过度预测极少数氨基酸（如酪氨酸和甘氨酸），同时忽略功能重要残基。我们将这一失败追溯到GNN编码器从有限结构数据从头学习氨基酸分布，丢弃进化数据库中编码的替换模式。为解决此问题，我们提出EvoStruct，通过交叉注意力适配器桥接冻结的蛋白质语言模型(PLM)与E(3)-等变GNN的3D结构上下文。与先前用于通用蛋白质设计的PLM-结构适配器不同，EvoStruct通过渐进PLM解冻和R-Drop一致性正则化针对CDR设计特有的词汇崩溃问题。在CHIMERA-Bench数据集上，EvoStruct在多种抗体设计方法中实现最高氨基酸恢复率和最低困惑度，相对于最佳GNN基线序列恢复率提高16%、困惑度降低43%，同时恢复2.3倍更高的氨基酸多样性和与真实值最高的结合对相关性。

原文摘要

Equivariant graph neural network (GNN) methods for antibody complementarity-determining region (CDR) design achieve the highest sequence recovery but suffer from severe vocabulary collapse. The current best GNN methods over-predict very few amino acids, such as tyrosine and glycine, while ignoring functionally important residues. We trace this failure to GNN encoders learning amino acid distributions de novo from limited structural data, discarding substitution patterns encoded in evolutionary databases. To resolve this, we propose EvoStruct, which bridges a frozen protein language model (PLM) with 3D structural context from an E(3)-equivariant GNN via a cross-attention adapter. Unlike prior PLM-structure adapters for general protein design, EvoStruct targets the vocabulary collapse proble...

自动采集于 2026-05-22

#论文 #arXiv #ML #小凯