Machine learning models identify ferroptosis-related genes as potential diagnostic biomarkers for Alzheimer’s disease

• Published in: Frontiers in aging neuroscience Vol. 14; p. 994130
• Main Authors: Deng, Yanyao, Feng, Yanjin, Lv, Zhicheng, He, Jinli, Chen, Xun, Wang, Chen, Yuan, Mingyang, Xu, Ting, Gao, Wenzhe, Chen, Dongjie, Zhu, Hongwei, Hou, Deren
• Format: Journal Article
• Language: English
• Published: Lausanne Frontiers Research Foundation 28.09.2022; Frontiers Media S.A
• Subjects: Algorithms; Alzheimer's disease; Apoptosis; Autophagy; bioinformatics; Cell death; Correlation analysis; Datasets; diagnostic model; Feature selection; Ferroptosis; FOXO1 protein; Gene expression; Gene set enrichment analysis; Inflammation; IQGAP1 protein; Learning algorithms; Lipid peroxidation; Lipids; Machine learning; machine learning algorithms; Neurodegeneration; Neurodegenerative diseases; Neuroscience; Oxidative stress; Pathogenesis; Toxicity
• ISSN: 1663-4365; 1663-4365
• DOI: 10.3389/fnagi.2022.994130

Alzheimer’s disease (AD) is a complex, and multifactorial neurodegenerative disease. Previous studies have revealed that oxidative stress, synaptic toxicity, autophagy, and neuroinflammation play crucial roles in the progress of AD, however, its pathogenesis is still unclear. Recent researches have indicated that ferroptosis, an iron-dependent programmed cell death, might be involved in the pathogenesis of AD. Therefore, we aim to screen correlative ferroptosis-related genes (FRGs) in the progress of AD to clarify insights into the diagnostic value. Interestingly, we identified eight FRGs were significantly differentially expressed in AD patients. 10,044 differentially expressed genes (DEGs) were finally identified by differential expression analysis. The following step was investigating the function of DEGs using gene set enrichment analysis (GSEA). Weight gene correlation analysis was performed to explore ten modules and 104 hub genes. Subsequently, based on machine learning algorithms, we constructed diagnostic classifiers to select characteristic genes. Through the multivariable logistic regression analysis, five features (RAF1, NFKBIA, MOV10L1, IQGAP1, FOXO1) were then validated, which composed a diagnostic model of AD. Thus, our findings not only developed genetic diagnostics strategy, but set a direction for further study of the disease pathogenesis and therapy targets.