Research On Methodology Of Mining SELDI Proteinchip Tumor Markers

It is well known that malignant tumors have been seriously threatening human's life and health. Early diagnosis and prevention on cancers plays an important role in tumor therapy. At present, Surface Enhanced Laser Desorption/Ionization Time of Flight Mass Spectrometry (SELDI-TOF-MS), a high-throughput and fast technology, has been generally used in clinical diagnosis of cancers.However, the reliability and the reproducibility of SELDI-TOF-MS technology are still relatively deficiency. Meanwhile, this high-throughput proteomics technology could generate massive data. This thesis proposed a kind of tumor markers mining method to extract tumor markers with high sensitivity and specificity from SELDI serum proteinchips. The main contributions of this paper are listed as follows:(1) Applied the data preprocessing method to obtain the effective protein peak clusters. The improvements including section threshold wavelet denoising method and protein mass dynamic calibration method are put forward to obtain more reliable and reproducible peak clusters from original proteinchips.(2) Proposed the diagnostic model associating genetic algorithm (GA) with specific classifiers for tumor marker extraction. This paper adopted genetic algorithm to search the optimal combination of distinct peak clusters and made some improvement on specific parameter settings. Moreover, support vector machines (SVM), artificial neural network (ANN) and k nearest neighborhood (KNN) are implemented and improved as tumor markers classifiers.(3) Gave an application of extracting tumor markers from colorectal cancer serum samples using the data preprocessing method and diagnostic model above. The extracted result with high sensitivity and specificity showed the tumor markers extracting methodology has a great value of clinical practices.