Analysis Model Of Proteomics Mass Spectrometry Data For Some Tumor Protein And Its Application

In recent years, malignant tumor is a serious threat to human health. Looking forpotential tumors biomarkers plays a very important role in the early diagnosis andtreatment of the malignant tumors. The proteomics data analysis based onSELDI-TOF MS is one of the most powerful tools for the early diagnosis of cancer. Itcan be enormously valuable to identify markers or that can reflect the differencesbetween samples from mass spectrometry data. Based on SELDI-TOF MS, a novelmethod of feature extraction was proposed with high accuracy, sensitivity andspecificity. The detail wok is as follows.A list of effective protein peaks was extracted from proteomic mass spectrometryraw data using the data preprocessing, including resampling, baseline correction,normalization, smoothing, peak finding with wavelets denoising, and spectralalignment of profiles.Two approaches for graphical representation were proposed to transform massspectrometry data to sequences, one is amino acid sequence and the other is0-1sequence, which is taken as the markers of samples.The binary classification model was constructed based on the characteristics ofsamples. This paper adopted support vector machine (SVM) as the classifier toimplement and improve the accuracy of distinct peak clusters. Then the parameters ofthe model were optimized. And the result was compared with others to show thehigher accuracy.Appling the method of data preprocessing and the analysis model constructedabove to another ovarian dataset, we extend the original binary to multiclassclassifications.In this thesis, the analysis results showed that our method was not only with highersensitivity and specificity, but also applied to all of the mass spectrometry dataanalysis.