Application Of Pattern Recognition In Rapid Mass Spectrometry Analysis Of Complex Matrix Samples

With the continuous progress and development of computer technology and detection technology,the instrument data which we got is growing rapidly,and various kinds of databases are emerging in endlessly.How to find the characteristics and laws among things in a large amount of data is the main topic and research content that the contemporary scholars urgently need to solve.Pattern recognition is one of the best methods to analyze and mine the collected data,and has been widely used in many fields.In this paper,partial least squares linear discriminant analysis(PLS-LDA),random forest(RF),partial least squares(PLS),principal component analysis(PCA)and other pattern recognition methods were used to analyze the mass spectrum data of lung cancer,water quality and tea,respectively.At the same time,the corresponding classification and recognition model is established,and good results are obtained,which provides a new research idea for the application of pattern recognition in mass spectrum data.The main contents of this paper are as follows:(1)A diagnosis model of lung cancer based on partial least squares linear discriminant analysis(PLS-LDA)was established.The direct mass spectrometry analysis of lung cancer tissue and normal tissue was carried out by electrospray ionization mass spectrometry(EIionization mass spectrometry),and the pattern of mass spectrometry data of lung cancer tissue and normal tissue was analyzed by partial least squares linear discriminant analysis(PLS-LDA).The potential biomarkers in tissue samples were identified and analyzed in order to further explore the occurrence and growth of lung cancer and to search for potential biomarkers laid the foundation.(2)Two direct mass spectrometry classification models of water quality based on random forest RF was established.The five kinds of surface water were analyzed by direct mass spectrometry,and the random forest algorithm was used to discriminate the five kinds of water quality samples.The accuracy of the model reached 95.19% and 100% respectively.(3)Establishing an algorithm based on random forest(RF)to rapidly implement classification of heavy metal-containing Cu water samples.The external accuracy of the model reached 96.15%.(4)Tea classification model based on partial least square method(PLS)and Random Forest(RF)was established.The partial least square method(PLS)and Random Forest(RF)has successfully realized the rapid classification of red tea and green tea and the rapid differentiation of tea from different producing areas.At the same time,it has also selected some potential marker ions,which is of great significance to the research and industry of tea classification.It also provides a new way for the classification of tea.