目的 开发基于机器学习的人类快速延迟整流性钾通道基因(the human Etherh-à-go-go-Reloted Gene,hERG)特异性打分函数(RF-hERG-Score),用于预测化合物对hERG的抑制活性(pIC50)。方法 采用随机森林算法,以AutoDock Vina(传统打分函数)分子对接生成的1 847个化合物-hERG复合体结构和实验测定的半抑制浓度(pIC50)数据作为训练集进行训练。结果 在十倍交叉验证中,RF-hERG-Score比RF-Score(通用打分函数)和AutoDock Vina更准确,其预测的pIC50与实验值的皮尔逊相关系数(Rp)为0.603、斯皮尔曼等级相关系数(Rs)为0.623、均方根误差(RMSE)为0.849。在两组外部测试集中,RF-hERG-Score的Rp、Rs和RMSE也高于其他2种方法,并且优于相应文献报道模型的预测性能。结论 RF-hERG-Score提高了hERG抑制剂结合亲和力的预测准确度,为利用计算模拟方法实现药物心脏毒性的较准确预测提供一种新的方案。
Abstract
OBJECTIVE To develop a machine learning-based hERG(the human Ether-à-go-go-Related Gene) target-specific scoring function(RF-hERG-Score) to predict the inhibitory activity of drugs on hERG(pIC50). METHODS The random forest algorithm was used, and the structures of 1847 compound-hERG complexes generated by AutoDock Vina molecular docking and experimental binding affinity(pIC50) data were used as the training set. RESULTS In ten-fold cross-validation, RF-hERG-Score was more accurate than RF-Score(generic scoring function) and AutoDock Vina(empirical scoring function). Between the pIC50 predicted by RF-hERG-Score and the experimental pIC50, the Pearson correlation coefficient(Rp) was 0.603, the Spearman rank correlation coefficient(Rs) was 0.623, and the root mean square error(RMSE) was 0.849. In the two external test sets, the Rp, Rs, and RMSE of RF-hERG-Score were also higher than the other two methods and better than the prediction performance of the model reported in the corresponding research. CONCLUSION RF-hERG-Score improves the prediction accuracy of the binding affinity of hERG inhibitors and provides a new solution for using computational simulation methods to achieve accurate prediction of drug cardiotoxicity.
关键词
人类快速延迟整流性钾通道基因 /
毒性预测 /
分子对接 /
机器学习 /
打分函数
{{custom_keyword}} /
Key words
hERG /
toxicity prediction /
molecular docking /
machine learning /
scoring function
{{custom_keyword}} /
中图分类号：
R911
{{custom_clc.code}}
({{custom_clc.text}})
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] BERGSTROM F, LINDMARK B. Accelerated drug discovery by rapid candidate drug identification[J]. Drug Discov Today,2019,24(6):1237-1241.
[2] VILLOUTREIX B O, TABOUREAU O. Computational investigations of hERG channel blockers: New insights and current predictive models[J]. Adv Drug Deliv Rev,2015,86:72-82. Doi:10.1016/j.addr.2015.03.003.
[3] SANGUINETTI M C, TRISTANI-FIROUZI M. hERG potassium channels and cardiac arrhythmia[J]. Nature,2006,440(7083):463-469.
[4] ZHOU S B, WANG J, LIU H. Lead compound optimization strategy(5) - reducing the hERG cardiac toxicity in drug development[J]. Acta Pharm Sin(药学学报), 2016, 51(10): 1530-1539.
[5] LIANG H. Classification of the human Ether-à-go-go-Related Gene channel blocking compounds[D], Beijing:Beijing University of Chemical Technology, 2011.
[6] ZHANG Y, ZHAO J, WANG Y, et al. Prediction of hERG K+ channel blockage using deep neural networks[J]. Chem Biol Drug Des,2019,94(5):1973-1985.
[7] WANG S, LI Y, XU L, et al. Recent developments in computational prediction of HERG blockage[J]. Curr Top Med Chem,2013,13(11):1317-1326.
[8] WANG S C. Studies on the computational prediction of hERG channel blockage[D], Suzhou:Soochow University, 2013.
[9] DU L P, YANG Q, YOU Q D, et al. In silico prediction of drug-hERG potassium channel interaction[J]. J China Pharm Univ(中国药科大学学报), 2006, 37(4): 291-296.
[10] SUN H, SCOTT D O. Structure-based drug metabolism predictions for drug design[J]. Chem Biol Drug Des,2010,75(1):3-17.
[11] WANG W, MACKINNON R. Cryo-EM structure of the open Human Ether-a-go-go-Related K(+) channel hERG[J]. Cell,2017,169(3):422-430.
[12] BUTLER A, HELLIWELL M V, ZHANG Y, et al. An update on the structure of hERG[J]. Front Pharmacol,2019,10:1572.Doi: 10.3389/fphar.2019.01572.
[13] LI H, SZE K H, LU G, et al. Machine-learning scoring functions for structure-based virtual screening[J]. WIREs Comput Mol Sci,2020,11(1):e1478.Doi: 10.1002/wcms.1478.
[14] LI H, SZE K H, LU G, et al. Machine-learning scoring functions for structure-based drug lead optimization[J]. WIREs Comput Mol Sci,2020,10(5):e1465. Doi:10.1002/wcms.1465.
[15] SHEN C, HU Y, WANG Z, et al. Beware of the generic machine learning-based scoring functions in structure-based virtual screening[J]. Brief Bioinform,2020,22(3):1-22.
[16] ZHANG L, AI H X, LI S M, et al. Virtual screening approach to identifying influenza virus neuraminidase inhibitors using molecular docking combined with machine-learning-based scoring function[J]. Oncotarget,2017,8(47): 83142-83154.
[17] YUAN S G, CHAN H C. Using PyMOL as a platform for computational drug design[J]. WIREs Comput Mol Sci,2017,7(2):e1298. Doi:10.1002/wcms.1298.
[18] GOODSELL D S, OLSON A J. Automated docking of substrates to proteins by simulated annealing[J]. Proteins Struct Funct Bioinf,1990,8(3):195-202.
[19] LIU M, ZHANG L, LI S, et al. Prediction of hERG potassium channel blockage using ensemble learning methods and molecular fingerprints[J]. Toxicol Lett,2020,332:88-96. Doi:10.1016/j.toxlet.2020.07.003.
[20] MUNAWAR S, VANDENBERG J I, JABEEN I. Molecular docking guided grid-independent descriptor analysis to probe the impact of water molecules on conformational changes of hERG inhibitors in drug trapping phenomenon[J]. Int J Mol Sci,2019,20(14):3385. Doi:10.3390/ijms20143385.
[21] TROTT O, OLSON A J. AutoDock Vina: improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading[J]. J Comput Chem, 2010, 31(2): 455-461.
[22] BALLESTER P J, MITCHELL J B. A machine learning approach to predicting protein-ligand binding affinity with applications to molecular docking[J]. Bioinformatics,2010,26(9):1169-1175.
[23] LI H, LEUNG K S, WONG M H, et al. Improving AutoDock Vina using random forest: the growing accuracy of binding affinity prediction by the effective exploitation of larger data sets[J]. Mol Inform,2015,34(2-3):115-126.
[24] LI H, LEUNG K S, WONG M H, et al. Improving AutoDock Vina using random forest: the growing accuracy of binding affinity prediction by the effective exploitation of larger data sets[J]. Mol Inform,2015,34(2-3): 115-126.
[25] BOUKHARTA L, KERANEN H, STARY-WEINZINGER A, et al. Computer simulations of structure-activity relationships for HERG channel blockers[J]. Biochemistry,2011,50(27):6146-6156.
[26] NEGAMI T, ARAKI M, OKUNO Y, et al. Calculation of absolute binding free energies between the hERG channel and structurally diverse drugs[J]. Sci Rep,2019,9(1):16586. Doi:10.1038/s41598-019-53120-6.
{{custom_fnGroup.title_cn}}
脚注
{{custom_fn.content}}
基金
国家自然科学基金资助项目资助(82003655);辽宁省科技厅重点研发计划项目资助(2019JH2/10300041);辽宁省教育厅科学研究经费项目资助(LQN201906)
{{custom_fund}}
PDF(2612 KB)
