OpenLB
Open Library of Bioscience
Advanced Search
Chemical science
Jan 19, 2022;13 (3): 816-833.

MGraphDTA: deep multiscale graph neural network for explainable drug-target binding affinity prediction.

Ziduo Yang, Weihe Zhong, Lu Zhao, Calvin Yu-Chian Chen
Author Information
  1. Ziduo Yang: Artificial Intelligence Medical Center, School of Intelligent Systems Engineering, Sun Yat-sen University Shenzhen 510275 China chenyuchian@mail.sysu.edu.cn +862039332153.
  2. Weihe Zhong: Artificial Intelligence Medical Center, School of Intelligent Systems Engineering, Sun Yat-sen University Shenzhen 510275 China chenyuchian@mail.sysu.edu.cn +862039332153.
  3. Lu Zhao: Artificial Intelligence Medical Center, School of Intelligent Systems Engineering, Sun Yat-sen University Shenzhen 510275 China chenyuchian@mail.sysu.edu.cn +862039332153.
  4. Calvin Yu-Chian Chen: Artificial Intelligence Medical Center, School of Intelligent Systems Engineering, Sun Yat-sen University Shenzhen 510275 China chenyuchian@mail.sysu.edu.cn +862039332153. ORCID
PMID: 35173947 DOI: 10.1039/d1sc05180f

Abstract

Predicting drug-target affinity (DTA) is beneficial for accelerating drug discovery. Graph neural networks (GNNs) have been widely used in DTA prediction. However, existing shallow GNNs are insufficient to capture the global structure of compounds. Besides, the interpretability of the graph-based DTA models highly relies on the graph attention mechanism, which can not reveal the global relationship between each atom of a molecule. In this study, we proposed a deep multiscale graph neural network based on chemical intuition for DTA prediction (MGraphDTA). We introduced a dense connection into the GNN and built a super-deep GNN with 27 graph convolutional layers to capture the local and global structure of the compound simultaneously. We also developed a novel visual explanation method, gradient-weighted affinity activation mapping (Grad-AAM), to analyze a deep learning model from the chemical perspective. We evaluated our approach using seven benchmark datasets and compared the proposed method to the state-of-the-art deep learning (DL) models. MGraphDTA outperforms other DL-based approaches significantly on various datasets. Moreover, we show that Grad-AAM creates explanations that are consistent with pharmacologists, which may help us gain chemical insights directly from data beyond human perception. These advantages demonstrate that the proposed method improves the generalization and interpretation capability of DTA prediction modeling.

References

Brief Bioinform. 2021 Mar 22;22(2):2141-2150 [PMID: 32367110]
J Biomed Inform. 2020 Oct;110:103547 [PMID: 32860883]
IEEE J Biomed Health Inform. 2021 Jun;25(6):1864-1872 [PMID: 33739926]
RSC Adv. 2020 Jun 1;10(35):20701-20712 [PMID: 35517730]
J Cheminform. 2017 Aug 14;9(1):45 [PMID: 29086168]
Brief Bioinform. 2021 Jan 18;22(1):451-462 [PMID: 31885041]
PLoS One. 2019 Aug 20;14(8):e0220113 [PMID: 31430292]
Methods. 2017 Oct 1;129:81-88 [PMID: 28549952]
J Comput Chem. 2013 May 5;34(12):1071-82 [PMID: 23299630]
J Chem Inf Model. 2021 May 24;61(5):2187-2197 [PMID: 33872000]
Brief Bioinform. 2021 May 20;22(3): [PMID: 32964234]
J Cheminform. 2020 Sep 1;12(1):51 [PMID: 33431044]
Brief Bioinform. 2015 Mar;16(2):325-37 [PMID: 24723570]
J Chem Inf Model. 2019 Oct 28;59(10):4131-4149 [PMID: 31580672]
Nucleic Acids Res. 2016 Jan 4;44(D1):D1202-13 [PMID: 26400175]
Bioinformatics. 2018 Sep 1;34(17):i821-i829 [PMID: 30423097]
Toxicol In Vitro. 1994 Oct;8(5):1053-60 [PMID: 20693071]
J Med Chem. 2005 Jan 13;48(1):312-20 [PMID: 15634026]
Biochim Biophys Acta. 2010 Mar;1801(3):289-98 [PMID: 19715772]
J Phys Chem Lett. 2021 May 6;12(17):4247-4261 [PMID: 33904745]
J Proteome Res. 2017 Apr 7;16(4):1401-1409 [PMID: 28264154]
Brief Bioinform. 2021 Jul 20;22(4): [PMID: 33147620]
J Comput Aided Mol Des. 2015 Sep;29(9):885-96 [PMID: 26201396]
PLoS Comput Biol. 2019 Jun 14;15(6):e1007129 [PMID: 31199797]
Chem Sci. 2018 Jun 6;9(24):5441-5451 [PMID: 30155234]
Bioinformatics. 2021 May 23;37(8):1140-1147 [PMID: 33119053]
IEEE Trans Neural Netw Learn Syst. 2018 Aug;29(8):3374-3387 [PMID: 28783645]
J Cheminform. 2017 Apr 18;9(1):24 [PMID: 29086119]
Chem Sci. 2017 Oct 31;9(2):513-530 [PMID: 29629118]
Nat Commun. 2021 Nov 22;12(1):6775 [PMID: 34811351]
Arch Pharm Res. 2016 Sep;39(9):1193-201 [PMID: 27387321]
Nat Struct Biol. 2002 Sep;9(9):646-52 [PMID: 12198485]
Nat Biotechnol. 2011 Oct 30;29(11):1046-51 [PMID: 22037378]
Nat Chem Biol. 2011 Apr;7(4):200-2 [PMID: 21336281]
Bioinformatics. 2008 Jul 1;24(13):i232-40 [PMID: 18586719]
Bioinformatics. 2019 Sep 15;35(18):3329-3338 [PMID: 30768156]
Bioinformatics. 2019 Jan 15;35(2):309-318 [PMID: 29982330]
Nucleic Acids Res. 2016 Jan 4;44(D1):D1045-53 [PMID: 26481362]
J Chem Inf Comput Sci. 1994 Jan-Feb;34(1):154-61 [PMID: 8144710]
J Chem Inf Model. 2021 Jan 25;61(1):46-66 [PMID: 33347301]
Bioinformatics. 2021 May 5;37(5):693-704 [PMID: 33067636]
Drug Discov Today Technol. 2019 Dec;32-33:89-98 [PMID: 33386099]
J Chem Inf Model. 2019 Mar 25;59(3):947-961 [PMID: 30835112]
Expert Opin Drug Metab Toxicol. 2005 Jun;1(1):91-142 [PMID: 16922655]
J Med Chem. 2021 May 27;64(10):6924-6936 [PMID: 33961429]
Chem Sci. 2020 Jan 8;11(9):2531-2557 [PMID: 33209251]
Brief Bioinform. 2021 Sep 2;22(5): [PMID: 33517357]
Drug Discov Today. 2016 Jan;21(1):82-89 [PMID: 26272035]
J Chem Inf Model. 2014 Mar 24;54(3):735-43 [PMID: 24521231]
Bioinformatics. 2018 Apr 1;34(7):1164-1173 [PMID: 29186331]
ACS Cent Sci. 2017 Apr 26;3(4):283-293 [PMID: 28470045]
Comput Struct Biotechnol J. 2013 Nov 11;8:e201308005 [PMID: 24688745]
Drug Discov Today. 2018 Jun;23(6):1241-1250 [PMID: 29366762]
J Chem Inf Model. 2020 Mar 23;60(3):1137-1145 [PMID: 31928003]
J Med Chem. 2020 Aug 27;63(16):8749-8760 [PMID: 31408336]
Bioinformatics. 2020 Aug 15;36(16):4406-4414 [PMID: 32428219]
Journal Article

Word Cloud

Similar Articles

Cited By