Prediction Of Escherichia Coli Resistant To Carbapenems Using Pattern Recognition

Escherichia Coli(E.coli)strains were selected as research samples,which was collected from the department of clinical laboratory of Xuzhou Medical University Affiliated Hospital.The peptides and metabolites extreacted from E.coli strains of carbapenem-resistant and carbapenemsensitive were analyzed by an orbitrap mass spectrometry(Orbitrap MS).Pattern recognition methods were introduced to realize rapid and accurate identification of drug-resistant and sensitive strains from complex MS data.The application of univariate statistical analysis provides a theoretical basis for studying the differences between drug-resistant and sensitive strains,and clarifies the mechanism of drug resistance.Realizing cluster analysis,principal component analysis(PCA),partial least squares discriminant analysis(PLS-DA)and orthogonal partial least squares discriminant analysis(OPLSDA)are "cluster" packages,"psych" package and "ropls" package,respectively.Univariate statistical analysis using the t test to calculate the P value of each compound,for significant differential screening.At the same time,the fold change(FC)analysis realize variation of the content of the compound in the resistant strain and the sensitive strain.Pattern recognition analysis of Orbitrap MS data from 24 peptide samples showed that clustering analysis based on the farthest distance method could well distinguish the two classes of E.coli,and the cophenetic correlation coefficient of the farthest distance method was 0.901,the highest one.The results of PCA,PLSDA and OPLS-DA were compared.The accuracy of OPLS-DA in identifying the two classes of E.coli strains was the highest.The differential compounds identified by PCA,PLS-DA and OPLSDA were analyzed by t test and fold changes(FC).PCA identified one compound that was significantly different in the two classes of strains.FC value of this compound was 1.5171.PLSDA identified 18 compounds that were significantly different in the two classes of strains.Among the 18 compounds,P values of 5 compounds were less than 0.01,indicating that differences of the 5 compounds were extremely significant.Only one compounds had a FC value less than 1.OPLSDA identified 26 compounds with significant differences in the two classes of strains.Among the 26 compounds,P values of 6 compounds were less than 0.01,and FC value of one compound was less than 1.The results of the cluster analysis based on the farthest distance method of the metabolite samples was the best,and the cophenetic correlation coefficient was 0.829,the highest one.PCA,PLS-DA and OPLS-DA of metabolite samples showed that PCA could not completely distinguish the two classes of strains,while PLS-DA and OPLS-DA could completely distinguish them.The predictive ability for two classes of strains were more than 90%.The results of t test and fold changes showed that 3 compounds with significant differences in the two classes of strains were screened by PCA,P values of 2 compounds were less than 0.01 and FC values of 3 compounds were less than 1.A total of 87 compounds with significant differences were screened out by PLS-DA.P value of 55 compounds were less than 0.01 and FC value of 7 compounds were more than 1.96 compounds with significant differences were screened out by OPLS-DA.P values of 70 compounds were less than 0.01 and FC values of 9 compounds were more than 1.