Zang Min, Gan Haipeng, Zhou Xuejie, Wang Lei, Dong Hao
Kuang Yaming Honors School, Nanjing University, Nanjing 210023, China.
State Key Laboratory of Natural Medicines and Jiangsu Key Laboratory of Drug Design and Optimization, and Department of Medicinal Chemistry, School of Pharmacy, China Pharmaceutical University, Nanjing 210009, China.
J Chem Inf Model. 2025 May 26;65(10):5113-5123. doi: 10.1021/acs.jcim.5c00629. Epub 2025 May 1.
Heat shock protein 90 (Hsp90) is a pivotal molecular chaperone crucial in the maturation of client proteins, positioning it as a significant target for cancer therapy. However, the design of effective Hsp90 inhibitors presents substantial challenges due to the complex interaction network and the requisite specificity of the inhibitors. This study tackles the task of designing peptide inhibitors capable of concurrently binding to both the ATP-binding pocket and the Cdc37-binding site within the N-terminal domain of Hsp90. In response to these challenges, we developed an advanced peptide screening protocol that merges machine learning with various molecular simulation techniques to boost the identification and optimization of potent inhibitors. Our integrated approach employs a convolutional neural network-based framework to predict peptide binding propensities. This predictive model is augmented by comprehensive molecular docking and dynamic simulations to assess the stability and interaction dynamics of Hsp90/peptide complexes. We successfully identified three heptapeptides that demonstrate the ability to interact with both binding sites, effectively obstructing the entrance to the ATP-binding pocket. This study elucidates the inhibitory mechanisms of these peptides, paves the way for the development of more efficacious therapeutic agents targeting Hsp90, and underscores the value of integrating machine learning techniques with molecular modeling in the peptide design process.
热休克蛋白90（Hsp90）是一种关键的分子伴侣，对客户蛋白的成熟至关重要，这使其成为癌症治疗的重要靶点。然而，由于复杂的相互作用网络以及抑制剂所需的特异性，设计有效的Hsp90抑制剂面临重大挑战。本研究致力于设计能够同时结合Hsp90 N端结构域内ATP结合口袋和Cdc37结合位点的肽抑制剂。针对这些挑战，我们开发了一种先进的肽筛选方案，该方案将机器学习与各种分子模拟技术相结合，以促进强效抑制剂的识别和优化。我们的综合方法采用基于卷积神经网络的框架来预测肽的结合倾向。通过全面的分子对接和动力学模拟对该预测模型进行增强，以评估Hsp90/肽复合物的稳定性和相互作用动力学。我们成功鉴定出三种七肽，它们能够与两个结合位点相互作用，有效阻碍ATP结合口袋的入口。本研究阐明了这些肽的抑制机制，为开发更有效的靶向Hsp90的治疗药物铺平了道路，并强调了在肽设计过程中将机器学习技术与分子建模相结合的价值。
