物理化学学报

上一篇    下一篇

BiCoA-Net：用于预测蛋白质-配体结合动力学的可解释双向协同注意力框架

李婧, 杨雨薇, 谷书凯, 王世航, 刘博, 刘焕香, 姚小军   

  1. 澳门理工大学应用科学学院, 人工智能药物发现中心, 澳门 999078
  • 收稿日期:2025-11-13 修回日期:2026-03-09 录用日期:2026-03-10
  • 通讯作者: 刘焕香, 姚小军 E-mail:hxliu@mpu.edu.mo;xjyao@mpu.edu.mo
  • 基金资助:
    本研究由澳门科学技术发展基金(0030/2024/RIA1)和澳门理工大学(RP/FCA-15/2023)资助

BiCoA-Net: an interpretable bidirectional co-attention framework for predicting protein-ligand binding kinetics

Jing Li, Yuwei Yang, Shukai Gu, Shihang Wang, Bo Liu, Huanxiang Liu, Xiaojun Yao   

  1. Center for Artificial Intelligence Driven Drug Discovery, Faculty of Applied Science, Macao Polytechnic University, Macao 999078, China
  • Received:2025-11-13 Revised:2026-03-09 Accepted:2026-03-10
  • Contact: Huanxiang Liu, Xiaojun Yao E-mail:hxliu@mpu.edu.mo;xjyao@mpu.edu.mo

PDF

43

赞

摘要： 定量理解蛋白质-配体结合动力学，特别是解离速率常数(koff)，对于表征分子识别动态和优化体内药效至关重要。然而，由于复杂的解离过渡态和持续存在的数据稀缺问题，从分子数据预测koff仍是一个核心挑战。本文通过构建KinetX (迄今为止最大的实验koff值综合基准数据集，涵盖256个不同蛋白质靶标的5624个测量值)来应对这些挑战。随后，我们提出BiCoA-Net——一个可解释的深度学习框架，旨在从序列表征中建模双向物理化学相互作用。该模型采用双向协同注意力机制和自适应门控融合层来学习这些复杂的非线性相互作用。BiCoA-Net实现了很好的预测精度，并在冷启动场景(即面对新型蛋白质靶标时)表现出强大的泛化能力。此外，该模型的可解释性特征可识别关键物理化学描述符，并揭示双模态相互作用策略。在四个临床相关保留靶标上的验证证实了其实用性，显示出稳健的化合物排序能力和虚拟筛选任务中的显著富集效果。BiCoA-Net与KinetX可以为药物-靶标结合动力学的计算预测与优化提供稳健、可泛化且可解释的框架。

关键词: 蛋白质-配体结合动力学, 解离速率常数, 深度学习, 双向共注意力网络, 可解释人工智能, 理性药物设计

Abstract: A quantitative understanding of protein-ligand binding kinetics, particularly the dissociation rate constant (koff), is fundamental to characterizing the dynamics of molecular recognition and optimizing in vivo pharmacodynamics. However, predicting koff from molecular data remains a central challenge due to complex unbinding transition states and persistent data scarcity. Here, we confront these challenges by first curating KinetX, the largest comprehensive benchmark dataset of experimental koff values to date, comprising 5624 measurements across 256 distinct protein targets. We then introduce BiCoA-Net, an interpretable deep learning framework designed to model reciprocal physicochemical interactions from sequence representations. The model employs a bidirectional co-attention mechanism and an adaptive gated fusion layer to learn these complex, non-linear interactions. BiCoA-Net achieves state-of-the-art accuracy and shows strong generalization in cold-start scenarios, where protein targets are novel. Furthermore, the model’s interpretability features identify key physicochemical descriptors and reveal bimodal interaction strategies. Validation on four held-out, clinically relevant targets confirms its practical utility, demonstrating robust compound ranking and significant enrichment in virtual screening tasks. BiCoA-Net and KinetX provide a robust, generalizable, and interpretable framework for the computational prediction and optimization of drug-target binding kinetics.

Key words: Protein-ligand binding kinetics, Dissociation rate constant, Deep learning, Bidirectional co-attention networks, Interpretable AI, Rational drug design