基于深度学习的中药抗菌组分分析及其在抗肺炎克雷伯菌感染组分中药开发中的应用  

Analysis of antimicrobial components of traditional Chinese medicine based on deep learning and its application in the development of antimicrobial component-based Chinese medicine against Klebsiella pneumonia

作　　者：徐吟秋 梁培[1,2] XU Yin-qiu;LIANG Pei(Department of Pharmacy,Nanjing Drum Tower Hospital,Affiliated to Medical School of Nanjing University,Jiangsu Nanjing 210008,China;Nanjing Medical Center for Clinical Pharmacy,Nanjing Drum Tower Hospital,Affiliated to Medical School of Nanjing University,Jiangsu Nanjing 210008,China)

机构地区：[1]南京大学医学院附属鼓楼医院药学部,江苏南京210008 [2]南京大学医学院附属鼓楼医院南京临床药学中心,江苏南京210008

出　　处：《中国医院药学杂志》2023年第10期1084-1088,共5页Chinese Journal of Hospital Pharmacy

基　　金：南京鼓楼医院药学部科研专项(编号:No.24)。

摘　　要：目的:以临床常见的致病菌肺炎克雷伯菌为对象,借助深度学习技术分析中药潜在抗菌组分,进而为开发抗菌组分中药提供参考。方法:针对肺炎克雷伯菌,以化学信息学数据库中收录的分子组成训练集与测试集,在借助深度神经网络构建深度学习模型后,对中药系统药理学数据库与分析平台(Chinese medicine systems pharmacology database and analysis platform,TCMSP)中各味中药所对应的中药单体展开抗菌活性预测,进而分析中药潜在抗菌组分。结果:经过超参数优化与训练,构建的深度学习模型对于测试集的正确率、精度、召回率、F1值分别为95.8%、96%、96.7%、0.963,曲线下面积值为0.991,该模型预测18味中药含抗肺炎克雷伯菌组分比例在20%以上,其中9味具有清热功效,同时大血藤、翻白草、金樱子、南五味子、商陆的抗菌活性已得到证实。结论:以肺炎克雷伯菌为对象,基于深度学习的中药抗菌组分预测模型,能够精准地识别中药抗菌组分,进而为开发抗菌组分中药提供参考。OBJECTIVE Klebsiella pneumoniae is a common pathogen in clinic.Deep learning technology is used to analyse the potential antibacterial components of traditional Chinese medicine,thereby providing reference for the development of antibacterial components of traditional Chinese medicine.METHODS Molecules were downloaded from cheminformatics databases to prepare a training set and a test set,which were used to construct a deep learning model based on deep neural networks.Then,the antimicrobial activity of the corresponding monomers in the traditional Chinese medicine systems pharmacology database and analysis platform(TCMSP)was predicted,followed by the analysis of potential antimicrobial components of CMM.RESULTS After hyperparameter optimization and training,the constructed deep learning model was tested with its accuracy,precision,recall and F1score of 95.8%,96%,96.7%and 0.963 respectively,and its area under the curve was 0.991.Eighteen herbs were found with antimicrobial components over 20%,and nine herbs had an effect of refrigeration.Finally,against Klebsiella pneumonia,the antimicrobial effects of Sargentodoxae Caulis,Potentillae Discoloris Herba,Rosae Laevigatae Fructus,Schisandrae Sphenantherae Fructus and Phytolaccae Radix were proven.CONCLUSION With Klebsiella pneumoniae as the target,the antimicrobial components prediction model based on deep learning made a precise prediction,which may further improve the development of antimicrobial component-based Chinese medicine.

关 键 词：深度学习 肺炎克雷伯菌 组分中药 

分 类 号：R91[医药卫生—药学]