Abstract
There is still a lack of effective therapies for Alzheimer's disease (AD), the leading cause of dementia and cognitive decline. Identifying reliable biomarkers and therapeutic targets is crucial for advancing AD research. In this study, we developed an aggregative multi-filter gene selection approach to identify AD biomarkers. This method integrates hub gene ranking techniques, such as degree and bottleneck, with feature selection algorithms, including Random Forest and Double Input Symmetrical Relevance, and applies ranking aggregation to improve accuracy and robustness. Five publicly available AD-related microarray datasets (GSE48350, GSE36980, GSE132903, GSE118553, and GSE5281), covering diverse brain regions like the hippocampus and frontal cortex, were analyzed, yielding 803 overlapping differentially expressed genes from 464 AD and 492 normal cases. An independent dataset (GSE109887) was used for external validation. The approach identified 50 prioritized genes, achieving an AUC of 86.8 in logistic regression on the validation dataset, highlighting their predictive value. Pathway analysis revealed involvement in critical biological processes such as synaptic vesicle cycles, neurodegeneration, and cognitive function. These findings provide insights into potential therapeutic targets for AD.