摘要
目的基于基质辅助激光解吸电离飞行时间质谱仪(MALDI-TOF MS)鉴定数据,用随机森林算法建立碳青霉烯耐药肺炎克雷伯菌(CR-KP)的快速分类检测模型。方法回顾性收集2022年6月至2023年12月浙江省荣军医院检测到的非重复肺炎克雷伯菌240株,同时收集了经MALDI-TOF MS鉴定且成功率90%以上的质谱峰图。根据药敏检测结果分为CR-KP 80株、碳青霉烯敏感肺炎克雷伯菌(CS-KP)160株。用K-means对质谱数据进行聚类并确定“质心”分类峰作为特征,构建随机森林分类模型,数据的70%作为训练集,30%作为测试集。分别用测试集分数、十倍交叉验证和ROC曲线评估模型的检测效能。结果构建的随机森林分类模型在测试集上的分数为0.94,十倍交叉验证分数为0.84,AUC为0.986(95%CI:0.950~1.000),特异度为0.980(95%CI:0.938~1.000),灵敏度为0.905(95%CI:0.814~0.995)。CR-KP和CS-KP质谱峰图的最大差异特征峰为4519 m/z。结论基于MALDI-TOF MS构建随机森林分类模型对识别CR-KP有较高的效能,可以为临床早期干预及抗生素的合理使用提供依据。
Objective To establish a rapid classification and detection model for carbapenem-resistant Klebsiella pneumoniae(CR-KP)by using the Random Forest algorithm based on matrix-assisted laser desorption ionization time-offlight mass spectrometry(MALDI-TOF MS).Methods From June 2022 to December 2023,a total of 240 cases of nonreplicated Klebsiella pneumoniae strains clinically detected in Zhejiang Rongjun Hospital were collected,and mass spectral peak maps identified by MALDI-TOF MS with a success rate of more than 90% were also collected.They were classified into 80 cases of CR-KP and 160 cases of carbapenem-sensitive Klebsiella pneumoniae(CS-KP)according to the results of the antimicrobial susceptibility testing.A random forest classification model was constructed by clustering the mass spectrometry data with K-means and identifying the"centre of mass"classification peaks as features,of which,70% of the data were used as the training set,while another 30% as the test set.The detection performance of the model was evaluated by test set scores,ten-fold cross validation and AUC of the subjects,respectively.Results The constructed random forest classification model had a score of 0.94 on the test set,a ten-fold cross validation score of 0.84,and an AUC of 0.986(95%CI:0.950-1.000),with a specificity of 0.980(95%CI:0.938-1.000),and a sensitivity of 0.905(95%CI:0.814-0.995).The maximum difference between CR-KP and CS-KP peaks was 4519 m/z.Conclusion The random forest classification model constructed based on MALDI-TOF MS shows high efficacy for identifying CR-KP,which can provide a basis for early clinical intervention and the rational use of antibiotics.
作者
许晓波
高芸涛
XU Xiaobo;GAO Yuntao(Department of Clinical Laboratory,Zhejiang Rongjun Hospital,Jiaxing 314000,China)
出处
《浙江医学》
2025年第4期387-391,共5页
Zhejiang Medical Journal
关键词
随机森林算法
机器学习模型
基质辅助激光解吸电离飞行时间质谱仪
肺炎克雷伯菌
Random forest algorithm
Machine learning model
Matrix-assisted laser desorption ionization time-offlight mass spectrometry
Klebsiella pneumoniae
