Neural Network And Wavelet Analysis Based Gene Expression Profile Dada Analysis

DNA microarrays can be used to measure the expression levels of thousandsof genes simultaneously and so it is usually an effective tool of cancer research.Gene expression data offer potential insight into gene function and regulatorymechanisms and aid in better understanding of carcinogenesis. DNAmicroarrays have been used in many research fields, such as cancerclassification and sub-classification, gene function and co-regulation researchand about how to produce more effective drugs, etc.Firstly, we discussed wavelet analysis based denoise method. It reaches agood denoising effective. And then we propose the joint use of discretewavelet transform and correlation coefficients ranking method to build afeature extractor. Some test and comparison experiments have been made toevaluate the performance of the proposed feature extraction scheme, and ithas been proved that: high classification rate, robustness, not sensitive to thenumber of features, and can be used in most classification system.After feature extraction, we research two application fields: cancerclassification and gene clustering. In cancer classification, we studiedprobabilistic neural network (PNN) mainly. We evaluate the performance usingPNN on three public data sets: Golub's Acute Leukemia data set,Bhattacharjee's adenocarcinoma data set and Alon's colon cancer data set.Correct rate higher than most of traditional methods has been obtained.The task of discovering every gene's functionality may not be as daunting as itsounds due to the surprising fact that there is a high level of similarity betweenthe genes and chromosomes of what were previously thought to be verydifferent organisms. We make use of clustering method to assist the researchabout gene function and co-regulated genes.