Data and code for "Delineating SARS-CoV-2 spike protein and antibodies interaction interfaces via siamese neural networks: a geometric and image-based analysis"

This archive contains the data and scripts used to (i) build homology-aware train/val/test splits at CDR-H3 level, (ii) render interface depth-map images (with optional augmentations) and compute geometric descriptors, and (iii) train/evaluate a cosine-contrastive Siamese CNN (two towers + fusion module). Please refer to 'README.txt' for details on the code.本归档文件包含用于完成以下三项任务的数据与脚本：(i) 在抗体重链第三互补决定区（CDR-H3）层面构建同源性感知的训练/验证/测试集划分方案；(ii) 渲染界面深度图（支持可选的数据增强操作）并计算几何描述符；(iii) 训练并评估余弦对比孪生卷积神经网络（双分支网络+融合模块）。 请参阅‘README.txt’以获取代码的详细说明。