[1]鲁　帅,尹帅领,原梦超,等.融合注意力机制和多尺度信息的蛋白质结合位点预测[J].郑州大学学报(工学版),2026,47(01):66-72.[doi:10.13705/j.issn.1671-6833.2026.01.008]
　LU Shuai,YIN Shuailing,YUAN Mengchao,et al.Protein Binding Site Prediction by Integrating Attention Mechanism and Multi-scale Information[J].Journal of Zhengzhou University (Engineering Science),2026,47(01):66-72.[doi:10.13705/j.issn.1671-6833.2026.01.008]
点击复制

融合注意力机制和多尺度信息的蛋白质结合位点预测()
分享到：

《郑州大学学报(工学版)》[ISSN:1671-6833/CN:41-1339/T]

卷:
47
期数:
2026年01期
页码:
66-72
栏目:
出版日期:
2026-01-06

文章信息/Info

Title:
Protein Binding Site Prediction by Integrating Attention Mechanism and Multi-scale Information
文章编号:
1671-6833(2026)01-0066-07
作者:
鲁　帅1; 2;  尹帅领3;  原梦超1; 2;  吴　迪1; 2;  周清雷1; 2; 3
1.郑州大学 计算机与人工智能学院,河南 郑州 450001;2.郑州大学 国家超级计算郑州中心,河南 郑州 450001;3.郑州大学 网络空间安全学院,河南 郑州 450002
Author(s):
LU Shuai1; 2;  YIN Shuailing3;  YUAN Mengchao1; 2;  WU Di1; 2;  ZHOU Qinglei1; 2; 3
1.School of Computer Science and Artificial Intelligence, Zhengzhou University, Zhengzhou 450001, China; 2.National Supercomputing Center in Zhengzhou, Zhengzhou University, Zhengzhou 450001, China; 3.School of Cyber Science and Engineering, Zhengzhou University, Zhengzhou 450002, China
关键词:
蛋白质结合位点预测;  3D U-Net;  压缩注意力机制;  多尺度信息;  噪声干扰
Keywords:
protein binding site prediction;  3D U-Net;  squeezed attention mechanism;  multi-scale information;  noise interference
分类号:
TP391Q-332
DOI:
10.13705/j.issn.1671-6833.2026.01.008
文献标志码:
A
摘要:
为了有效解决3D U-Net在蛋白质结合位点预测中存在的噪声干扰和多尺度信息缺乏问题,提出了一种融合注意力机制和多尺度信息的蛋白质结合位点预测模型AMPocket。引入压缩注意力机制,使得模型能够聚焦于关键通道的蛋白质特征,减少无关通道特征对结合位点预测的影响,从而提高分割的精度;在编码器中引入多尺度信息,使模型能够从不同层次捕捉特征,进而获得更加全面和丰富的空间信息。实验结果表明:AMPocket在4个广泛使用的测试集上均取得了优异的预测结果,特别是在SC6K数据集上的DCC成功率和DVO优于所有对比方法,分别为93.04%和55.01%,并且AMPocket具有较少的参数量,表明模型具有更好的预测性能。
Abstract:
To effectively address the issues of noise interference and insufficient multi-scale information within 3D U-Net for protein binding site prediction, a novel model named AMPocket was proposed which incorporated both attention mechanisms and multi-scale information to improve the accuracy of binding site prediction. AMPocket initially employed squeezed attention mechanism that enabled the model to focus on the most critical channels of protein features while diminishing the impact of irrelevant channels, thereby enhancing segmentation accuracy. Additionally, the multi-scale information was introduced to the encoder component, allowing the model to capture feature representations at various levels and thus obtained more comprehensive and robust spatial information. The experimental results demonstrated that AMPocket achieved superior predictive performance across four widely used test sets, in particular, the DCC success rate and DVO metrics on the SC6K dataset outperformed all other competing methods by 93.04% and 55.01% respectively, with a smaller number of parameters. It indicated that the model had better predictive performance.

参考文献/References:

[1]XIA Y, PAN X Y, SHEN H B. A comprehensive survey on protein-ligand binding site prediction[J]. Current Opinion in Structural Biology, 2024, 86: 102793.

[2]LE GUILLOUX V, SCHMIDTKE P, TUFFERY P. Fpocket: an open source platform for ligand pocket detection[J]. BMC Bioinformatics, 2009, 10: 168.
[3]BINKOWSKI T A, NAGHIBZADEH S, LIANG J. CASTp: computed atlas of surface topography of proteins[J]. Nucleic Acids Research, 2003, 31(13): 3352-3355.
[4]SIGRIST C J A, DE CASTRO E, CERUTTI L, et al. New and continuing developments at PROSITE[J]. Nucleic Acids Research, 2013, 41: D344-D347.
[5]KRIVÁK R, HOKSZA D. P2Rank: machine learning based tool for rapid and accurate prediction of ligand binding sites from protein structure[J]. Journal of Cheminformatics, 2018, 10(1): 39.
[6]INNIS C A. siteFiNDER|3D: a web-based tool for predicting the location of functional sites in proteins[J]. Nucleic Acids Research, 2007, 35: W489-W494.
[7]ZHOU H Y, CAO H N, SKOLNICK J. FINDSITEcomb2.0: a new approach for virtual ligand screening of proteins and virtual target screening of biomolecules[J]. Journal of Chemical Information and Modeling, 2018, 58(11): 2343-2354.
[8]SCHYMKOWITZ J, BORG J, STRICHER F, et al. The FoldX web server: an online force field[J]. Nucleic Acids Research, 2005, 33: W382-W388.
[9]RAVINDRANATH P A, SANNER M F. AutoSite: an automated approach for pseudo-ligands prediction-from ligand-binding sites identification to predicting key ligand atoms[J]. Bioinformatics, 2016, 32(20): 3142-3149.
[10] FRIESNER R A, BANKS J L, MURPHY R B, et al. Glide: a new approach for rapid, accurate docking and scoring. 1. method and assessment of docking accuracy[J]. Journal of Medicinal Chemistry, 2004, 47(7): 1739-1749.
[11]池燕飞,李春,冯旭东.机器学习在蛋白质功能预测领域的研究进展[J].生物工程学报,2023,39(6):2141-2157.
CHI Y F, LI C, FENG X D. Advances in machine learning for predicting protein functions[J]. Chinese Journal of Biotechnology, 2023, 39(6): 2141-2157.
[12] KOZLOVSKII I, POPOV P. Computational methods for binding site prediction on macromolecules[J]. Quarterly Reviews of Biophysics, 2025, 58: 1-31.
[13] RONNEBERGER O, FISCHER P, BROX T. U-Net: convolutional networks for biomedical image segmentation[C]∥ Medical Image Computing and Computer Assisted Intervention-MICCAI 2015. Cham: Springer, 2015: 234-241.
[14] ÇIÇEK Ö, ABDULKADIR A, LIENKAMP S S, et al. 3D U-Net: learning dense volumetric segmentation from sparse annotation[J]. Lecture Notes in Computer Science, 2016, 9901: 424-432.
[15] JIMÉNEZ J, DOERR S, MARTÍNEZ-ROSELL G, et al. DeepSite: protein-binding site predictor using 3D-convolutional neural networks[J]. Bioinformatics, 2017, 33(19): 3036-3042.
[16] MYLONAS S K, AXENOPOULOS A, DARAS P. DeepSurf: a surface-based deep learning approach for the prediction of ligand binding sites on proteins[J]. Bioinformatics, 2021, 37(12): 1681-1690.
[17] STEPNIEWSKA-DZIUBINSKA M M, ZIELENKIEWICZ P, SIEDLECKI P. Improving detection of protein-ligand binding sites with 3D segmentation[J]. Scientific Reports, 2020, 10(1): 5035.
[18] AGGARWAL R, GUPTA A, CHELUR V, et al. DeepPocket: ligand binding site detection and segmentation using 3D convolutional neural networks[J]. Journal of Chemical Information and Modeling, 2022, 62(21): 5069-5079.
[19] LI P Y, CAO B H, TU S K, et al. RecurPocket: recurrent Lmser network with gating mechanism for protein binding site detection[C]∥2022 IEEE International Conference on Bioinformatics and Biomedicine (BIBM). Piscataway: IEEE, 2022: 334-339.
[20] LI P Y, LIU Y C, TU S K, et al. GLPocket: a multi-scale representation learning approach for protein binding site prediction[C]∥ IJCAI International Joint Conference on Artificial Intelligence. Macao: JCAI,2023: 4821-4828.
[21] LIN X L, YANG W L, CHEN Y Y, et al. ResPocket: a multi-scale feature fusion method for improving protein binding site detection[C]∥2024 IEEE International Conference on Bioinformatics and Biomedicine (BIBM). Piscataway: IEEE, 2024: 1588-1591.
[22] HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C]∥2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2018: 7132-7141.
[23] KELLENBERGER E, MULLER P, SCHALON C, et al. sc-PDB: an annotated database of druggable binding sites from the Protein Data Bank[J]. Journal of Chemical Information and Modeling, 2006, 46(2): 717-727.
[24]WANG R X, FANG X L, LU Y P, et al. The PDBbind database: methodologies and updates[J]. Journal of Medicinal Chemistry, 2005, 48(12): 4111-4119.

更新日期/Last Update: 2026-01-17