| Microbial classification and identification technology is widely used in clinical,CDC,food,drug,entry-exit inspection and quarantine and other fields.Among them,clinical microbial identification plays an important role in diagnosis of infectious diseases,medication guidance,nosocomial infection control,antibacterial agents management,etc.The detection and identification of food-borne pathogens which can cause food poisoning such as E.coli O157:H7 and Salmonella enteritidis in meat and vegetables has become an indispensable key link in food hygiene industry.Meanwhile,accurate and rapid identification of microorganisms is indispensable in traceable monitoring of pathogenic bacteria in CDC,inspection and supervision of microbial contamination in drugs,and detection of pathogenic bacteria in import and export food,drugs and cosmetics.For now,the commonly used methods of microbial identification are mainly morphological observation and biochemical identification which based on traditional phenotype,and gene identification based on molecular genetics,such as 16S rRNA,PCR and so on.However,the methods above have their own disadvantages,which lead to the problems of long sample circulation time,complex operation,low identification accuracy and high cost.Matrix-assisted laser desorption ionization time-of-flight mass spectrometry(MALDI-TOF MS)is a kind of soft ionization time-of-flight mass spectrometry,which can be used for rapid,accurate and high throughput identification by using peptide mass fingerprinting(PMF).It's a revolution in microbial identification.Since the commercialized VITEK MS and BioTyper microbial mass spectrometry identification system were certified by FDA in 2012 and 2013,seperately,the technology has been widely used and evaluated in many fields in a few years because of the advantages above.The technical route of this paper is to obtain the peptide mass fingerprinting of microorganisms by using MALDI-TOF MS,and then,search and compare the obtained PMF with the established standard PMF database to complete the classification and identification of microorganisms.Based on the contents above,this paper is mainly consists of five chapters:In the first chapter,the advantages and significance of the subject are expounded by comparing with other classification and identification techniques in principle and characteristics.At present,microbial identification methods can be divided into two categories:traditional phenotypic identification and molecular genetic identification.It can be identified from four levels:1.the level of morphology,physiology and biochemistry of microorganisms;2.the level of microbial cell components;3.the level of microbial protein;4.the level of microbial genetic genes(nucleic acid).This chapter focuses on the principle,characteristics and process of PMF obtained by MALDI-TOF MS,which based on protein level,as well as the main frontier applications for now.In the second chapter,PMF acquisition strategy is optimized.First,the key structure and electrical parameters of ion source and mass analyzer are optimized from the point of quality(resolution)of spectrum.Two-field acceleration,delayed extraction(DE)and time-of-flight mass analyzer(TOF MS)theoretical models considering the effects of initial kinetic energy dispersion and spatial dispersion of ions are established.Based on this model,the focusing conditions are deduced,and a fast calculation method for solving the TOFMS model under the influence of multi-variables by using the Particle Swarm Optimization(PSO)is proposed to optimize the design.The optimization of several important parameters(5 parameters in this paper)of instruments can be completed simultaneously within a short time(a few minutes),which solves the problem that only one parameter can be optimized under the influence of multi-variable coupling without multi-variable optimization at the same time.Secondly,the influence of sample pretreatment and acquisition factors are analyzed.Fuzzy control technology is used and evaluated in automatic acquisition strategy for automatic laser energy adjustment.The quality and repeatability of the spectra are better than that of manual acquisition,and it is conducive to the standardization of data acquisition.In the third chapter,the spectral preprocessing algorithm is studied.Comparing and evaluating many commonly used algorithms in literature,such as smoothing,baseline correction,normalization,peak extraction,mass calibration and spectral alignment.A baseline correction algorithm combining median filter with wavelet transform,local minimum filter and polynomial fitting is designed,which solves the problem of over-fitting of baseline caused by wavelet transform and polynomial fitting.At the same time,a quadratic dynamic calibration algorithm is designed to realize the spectral alignment,which ensures the accuracy of the theoretical model of mass.In the fourth chapter,the construction of PMF database and classification and identification algorithm are studied.Designed the construction algorithm of feature weighted main spectrum prediction(MSP)and the retrieval and identification algorithm.48 strains of common clinical bacteria and fungi were identified by using Euclidean distance and Pearson correlation coefficient etc.The accuracy of the Feature-Weighted identification algorithm designed in this paper is 100%,and the average identification score is 2.43,which is better than other algorithms.An intelligent classification model based on receiver operating characteristic(ROC)curve and support vector machine(SVM)and a classification model based on back propagation neural network(BP-ANN)are established to realize fast and supervised classification of features.Two characteristic peaks(6571.92Da and 6701.61 Da)were extracted from ROC,and then fast and accurate classification was achieved to the accuracy of 96.77%by using linear SVM and four-layer BP neural network algorithm combined with typing research of 32 MRSA and 29 MSSA samples.In the fifth chapter,evaluation and verification are carried out for the above instrument and algorithm design through a large number of experiments.696 clinical samples including 46 genera and 85 species were tested in parallel with different methods.The difference comparison algorithm with feature weighting designed in the study achieved identification accuracy in genus and species was 99.86%and 94.68%,respectively.The average score is 2.37,which is obviously superior to other methods mentioned in the literature.Homology analysis of 29 nosocomial infections pathogens from different sources showed that the accuracy of hierarchical clustering(HCA)traceability was 96.55%.Thirdly,32 MRSAs and 29 MSSAs mentioned in Chapter 4 were used to unsupervised classification,and the accuracy rate was 88.52%.A new combined typing technology of MALDI-TOF MS and FT-IR is proposed.Through the typing test of easily confused Escherichia coli and Shigella,the accuracy of MALDI-TOF MS and FT-IR data was 68%and 89%,respectively.However,the accuracy of the combined typing technology was greatly improved,which reached 100%.In summary,this paper focuses on MALDI-TOF MS microbial PMF identification technology,analyses the influencing factors of PMF acquisition in detail,studies and optimizes the hardware structure,electrical parameters and key steps of acquisition,focuses on the design of classification and identification algorithms,and evaluates the above-mentioned work through a large number of different types of experiments and data.Completes the technology from data acquisition,spectral preprocessing,database construction and identification,finally carries out in application,finishing systematic research on key technologies.The results show that the MALDI-TOF MS microbial classification and identification algorithm designed in this paper is superior to the relevant methods in the existing references,and it has a high value for both algorithm research and practical application. |
PDF Full Text Request