Genetic information has recently attracted a significant attention in the diagnosis and In this thesis, by investigating different approaches of gene selection, a novel framework for gene selection is proposed, which uses the advantageous features of conventional methods and covers their weak points. In addition to gene expression values, the proposed method uses gene ontology, which is a reliable source of information on genes. Use of gene ontology beside gene expression data, can compensate in part for the limitations of microarrays; including having a small number of samples and erroneous measurement results. In the proposed framework, at first a significant number of irrelevant genes are omitted using the filtering method (fisher). Since filtering methods do not take into account the correlation among genes, the remaining genes will still have a large amount of redundancy. In order to reduce redundancy in remaining genes a greedy approach has been proposed for removing similar genes. This approach calculates the similarity between genes considering the gene ontology information as well as gene expression data using a hybrid criterion and then removes redundant genes according to this criterion. Finally genes that remain after this stage are processed more accurately by the SVMRFE method to derive the disease marker genes. The proposed method has been applied on DLBCL and colon cancer datasets. It is observed that the proposed method improves the performance of Microarray data sets often contain missing value due to different reasons including scratches or dust on the slide, error in experiments, image corruption and insufficient resolution. In this thesis a novel method is proposed which integrates CST clustering and gene ontology to estimate missing values at the preprocessing stage. The performance of the proposed method has been studied on the DLBCL data sets with different percentage of missing values. Comparing the results of proposed method with other existing estimating methods shows that the proposed method can estimate missing values with a higher accuracy. Key Words Gene Selection, Gene Ontology, Gene Expression, Microarray, Missing Value
