基于深度神经网络的DYRK1A抑制剂预测模型研究

中国医药导刊 ›› 2022, Vol. 24 ›› Issue (7): 672-677.

基于深度神经网络的DYRK1A抑制剂预测模型研究

王亚铃1，2，钱晨亮1，姚明丽1，司鑫鑫1*，夏杰2*

1.江苏海洋大学药学院，江苏连云港 222005； 2.中国医学科学院北京协和医学院药物研究所，北京 100050

收稿日期:2022-06-14 修回日期:2022-01-20 出版日期:2022-07-28 发布日期:2022-07-28
基金资助:
中国医学科学院医学与健康科技创新工程重大协同创新项目（项目编号：2021-I2M-1-069；项目名称：基于人工智能的靶向药物发现技术研究）

Deep Neural Network-based Models to Predict DYRK1A Inhibitors

1.School of Pharmacy， Jiangsu Ocean University， Jiangsu Lianyungang 222005， China； 2.Institue of Materia Medica， Chinese Academy of Medical Sciences & Peking Union Medical College，Beijing 100050， China

Received:2022-06-14 Revised:2022-01-20 Online:2022-07-28 Published:2022-07-28

摘要/Abstract

摘要： 目的：构建基于深度神经网络（deep neural network， DNN）的双底物特异性酪氨酸磷酸化调节激酶1A（DYRK1A）抑制剂预测模型，为筛选DYRK1A抑制剂提供计算工具。方法：从ChEMBL数据库中收集了DYRK1A抑制剂和非抑制剂共927个，通过随机采样10次建立了10组训练集和测试集，并计算出每个化合物的2种不同的分子特征，即MACCS指纹和Morgan2指纹，采用深度神经网络算法建立了20个模型，通过比较研究确定其中性能最佳的模型，并且对模型进行了Y-随机化检验、应用域和相似度图分析。结果：基于第2对训练-测试集Morgan2指纹的DNN模型（DNN1_Morgan2）性能最佳，对测试集内化合物的分类准确度为0.821，马修斯相关系数和ROC曲线下面积分别为0.647和0.917。结论：此最优模型可用于 DYRK1A抑制剂的活性预测、虚拟筛选，并用于先导化合物的设计及优化。

关键词: font-size:medium, ">双底物特异性酪氨酸磷酸化调节激酶1A；机器学习；预测模型

Abstract: Objective：This study is aimed to build a deep neural network(DNN)-based model to predict dual specificity tyrosine phosphorylation-regulated kinase 1A(DYRK1A) inhibitors， so as to provide a computational tool for virtual screening of DYRK1A inhibitors. Methods：A total of 927 DYRK1A inhibitors and non-inhibitors were collected from the ChEMBL database and 10 pairs of training sets and test sets were generated by random sampling for 10 times. Two different molecular features， namely MACCS fingerprints and Morgan2 fingerprints， were calculated for each compound， and then 20 models were built with the DNN algorithm. The best-performing model was determined by the comparative study and was further studied by the Y-randomization tests， applicability domain and similarity maps analysis. Results：The DNN model based on the 2nd pair of training and test sets and Morgan2 fingerprints(namely DNN1_Morgan2)， showed the best performance.The accuracy of the optimal model on the test set was 0.821. The Matthews correlation coefficient and the ROC AUC value were 0.647 and 0.917， respectively. Conclusion：The optimal model can be used for activity prediction and virtual screening of DYRK1A inhibitors， as well as lead compounds design and optimization.

Key words: font-size:medium, ">Dual specificity tyrosine phosphorylation-regulated kinase 1A； Machine learning； Prediction model

中图分类号:

王亚铃, 钱晨亮, 姚明丽, 司鑫鑫, 夏杰. 基于深度神经网络的DYRK1A抑制剂预测模型研究[J]. 中国医药导刊, 2022, 24(7): 672-677.

WANG Yaling, QIAN Chenliang, YAO Mingli, SI Xinxin, XIA Jie. Deep Neural Network-based Models to Predict DYRK1A Inhibitors[J]. CHINESE JOURNAL OF MEDICINAL GUIDE, 2022, 24(7): 672-677.

参考文献

[1]Liu T，Wang Y，Wang J，et al.DYRK1A inhibitors for disease therapy：current status and perspectives[J]. Eur J Med Chem， 2022， 229：114062.
     [2]Rammohan M， Harris E， Bhansali RS， et al. The chromosome 21 kinase DYRK1A：emerging roles in cancer biology and potential as a therapeutic target[J].Oncogene，2022， 41(14)：2003-2011.
     [3]Jarhad DB，Mashelkar KK，Kim HR， et al. Dual-specificity tyrosine phosphorylation-regulated kinase 1A(DYRK1A) inhibitors as potential therapeutics[J].J Med Chem，2018，61(22)：9791-9810.
     [4]Schmitt C，Miralinaghi P，Mariano M， et al. Hydroxybenzothiophene ketones are efficient pre-mRNA splicing modulators due to dual inhibition of dyrk1A and Clk1/4[J].ACS Med Chem Lett，2014，5(9)：963-967.
     [5]Demuro S，Di Martino RMC，Ortega JA，et al.GSK-3beta， FYN， and DYRK1A：master regulators in neurodegenerative pathways[J]. Int J Mol Sci，2021，22(16)：9098.
     [6]Laham AJ，Saber-Ayad M，El-Awady R.DYRK1A：a down syndrome-related dual protein kinase with a versatile role in tumorigenesis[J]. Cell Mol Life Sci， 2021， 78(2)：603-619.
     [7]Stotani S，Giordanetto F，Medda F. DYRK1A inhibition as potential treatment for Alzheimer′s disease[J]. Future Med Chem， 2016， 8(6)：681-696.
     [8]Fernandez-Martinez P，Zahonero C，Sanchez-Gomez P.DYRK1A：the double-edged kinase as a protagonist in cell growth and tumorigenesis[J].Mol Cell Oncol，2015，2(1)：e970048.
     [9]Kumar K，Suebsuwong C，Wang P，et al.DYRK1A inhibitors as potential therapeutics for beta-Cell regeneration for diabetes[J]. J Med Chem，2021，64(6)：2901-2922.
     [10]Wang D，Wang F，Tan Y，et al.Discovery of potent small molecule inhibitors of DYRK1A by structure-based virtual screening and bioassay[J]. Bioorg Med Chem Lett， 2012， 22(1)：168-171.
     [11]Kumar K，Man-Un Ung P，Wang P，et al.Novel selective thiadiazine DYRK1A inhibitor lead scaffold with human pancreatic beta-cell proliferation activity[J].Eur J Med Chem，2018，157：1005-1016.
     [12]Koyama T，Yamaotsu N，Nakagome I，et al.Multi-step virtual screening to develop selective DYRK1A inhibitors[J].J Mol Graph Model，2017，72：229-239.
     [13]Choi M，Kim AK，Ham Y，et al.Aristolactam BIII，a naturally derived DYRK1A inhibitor， rescues Down syndrome-related phenotypes[J].Phytomedicine，2021， 92：153695.
     [14]Li S，Ding Y，Chen M，et al. HDAC3i-Finder：a machine learning-based computational tool to screen for HDAC3 inhibitors[J].Mol Inform，2021，40(3)：e2000105.
     [15]Moriwaki H，Tian YS，Kawashita N，et al.Mordred：a molecular descriptor calculator[J].J Cheminform，2018，10(1)：4.
     [16]Li Y，Tian Y，Qin Z，et al.Classification of HIV-1 protease inhibitors by machine learning methods[J].ACS Omega，2018，3(11)：15837-15849.
     [17]Escanilla NS，Hellerstein L，Kleiman R，et al.Recursive feature elimination by sensitivity testing [C]. 2018 17th IEEE International Conference on Machine Learning and Applications(ICMLA), IEEE， 2018.
     [18]Vogt M，Bajorath J.Modeling tanimoto similarity value distributions and predicting search results[J]. Mol Inform，2017，36(7)：1600131.
     [19]Morgan HL.The generation of a unique machine description for chemical structures-a technique developed at chemical abstracts service[J]. J Chem Doc，2002，5(2)：107-113.
     [20]Ma J，Sheridan RP，Liaw A，et al.Deep neural nets as a method for quantitative structure-activity relationships[J]. J Chem Inf Model， 2015， 55(2)：263-274.
     [21]Matthews BW. Comparison of the predicted and observed secondary structure of T4 phage lysozyme[J]. Biochim Biophys Acta， 1975， 405(2)：442-451.
     [22]Hanley JA，Mcneil BJ.The meaning and use of the area under a receiver operating characteristic(ROC) curve[J]. Radiology， 1982， 143(1)：29-36.
     [23]Tang H，Wang XS，Huang XP，et al.Novel inhibitors of human histone deacetylase(HDAC) identified by QSAR modeling of known inhibitors，virtual screening，and experimental validation[J].J Chem Inf Model， 2009， 49(2)：461-476.
     [24]Riniker S，Landrum GA.Open-source platform to benchmark fingerprints for ligand-based virtual screening[J]. J Cheminform， 2013， 5(1)：26.

快速检索引用检索图表检索高级检索