The Research Of Feature Selection Algorithm Based On Swarm Intelligence In SELDI Mass Spectral Data Analysis

Applying feature selection (FS) techniques in bioinformatics has become a real prerequisite for model building. In particular, the high dimensional and small sample sizes natures of many modeling tasks in bioinformatics, going from sequence analysis over microarray analysis to spectral analyses and literature mining has given rise to a wealth of feature selection techniques being presented in the field. Small sample sizes and their inherent risk of imprecision and overfitting pose a great challenge for many modeling problems in bioinformatics. Specific applications in bioinformatics have led to a wealth of newly proposed techniques.Mass spectrometry (MS) technology is used to measure the mixture of proteins/peptides of biological tissues or fluids, such as serum or urine. Such measurements can be used to identify disease-related patterns, which hold potential for early diagnosis, prognosis, monitoring disease progression, response to treatment and drug target research.Comprehensive analyses on Surface-enhanced laser desorption/ionization time-of-flight mass spectrometry (SELDI-TOF-MS) data analyses are mainly discussed in our work and the application of swarm intelligence algorithm combined with SVM in biomarker selection is also studied in the work. The main contents of this dissertation are as follows:(1) The thesis researched fundamental principle of SELDI-TOF-MS technology and summarized various methods of its two main phases: pre-processing and biomarker selection. And its shortcomings and progress are discussed here.(2) Research on fundamental principle of Ant Colony Optimization Algorithm (ACO), Particle Swarm Optimization Algorithm (PSO) and their improved methods provides theoretical principles for further learning.(3) New method is raised using weighting factor as prior information in the ant colony optimization searching process. Combined with support vector machines (SVM), it was applied to identify relevant serum proteomic biomarkers. Experiments proposed method has strong power in distinguishing cancer patients from healthy individuals.(4) Combined SVM with QPSO, ACO and PSO, and using the models biomarkers selection, the experiments show that model built by QPSO achieved not only high prediction accuracy but also extremely fast velocity, so the proposed method QPSO-SVM has a certain good theoretical and utility value.The main contributions of this paper are summarized and the further researches on work are suggested at the end of this dissertation.