Exploring the Efficiency of Hub Genes in Identification of Alzheimer Disease

Alzheimer's disease (AD) is a kind of dementia that gets worse over time, it results in a loss in memory and cognitive function. Consequently, it is very important to diagnose AD in its early stages. This diagnosis is required using different methods through certain alternative approaches. Different techniques such as imaging, clinical, genetic, and fluid biomarker-based pattern classification methods are used for developing predictive models of AD dementia. In this paper, two approaches are used to recognize the prognostic and diagnostic biomarkers that can discriminate between AD and non-AD patients. The publicly available microarray of gene expression datasets from six brain regions are used. The first approach is used to explore the biomarker gene in each region of the six brain regions and in turns investigate the best region that used for AD identification. The second approach is to create the gene network based on extracting the most significant gene in each region separately and extract the hub one of each network. Then, hub genes in each region are used in a classification step to investigate the efficiency of such genes in recognizing AD patients.The highest 50 genes from each region were used in both approaches. In the classification step, the feature selection based on T-test followed by the Support Vector Machine (SVM) classifier is used. Experimental results show reliability of SVM with this kind of gene expression data. The first approach yields best classification results with Entorhinal Cortex (EC) region reaching 95.7%. The second approach proves that hub genes are more efficient in identification of AD improving the classification accuracy with all brain regions reaching 100% accuracy for EC region.