AUTHOR=Ma Wenhao , Su Yuelin , Zhang Peng , Wan Guoqing , Cheng Xiaoqin , Lu Changlian , Gu Xuefeng TITLE=Identification of mitochondrial-related genes as potential biomarkers for the subtyping and prediction of Alzheimer’s disease JOURNAL=Frontiers in Molecular Neuroscience VOLUME=Volume 16 - 2023 YEAR=2023 URL=https://www.frontiersin.org/journals/molecular-neuroscience/articles/10.3389/fnmol.2023.1205541 DOI=10.3389/fnmol.2023.1205541 ISSN=1662-5099 ABSTRACT=Alzheimer's disease (AD) is a progressive and debilitating neurodegenerative disorder prevalent among older adults. Although AD symptoms can be managed through certain treatments, advancing the understanding of underlying disease mechanisms and developing effective therapies is critical. In the present study, transcriptome data from the temporal lobes of healthy individuals and patients with AD were systematically analyzed to investigate the relationship between AD and mitochondrial autophagy. Machine learning algorithms were used to identify six genes-FUNDC1, MAP1LC3A, CSNK2A1, VDAC1, CSNK2B, and ATG5-for the construction of an AD prediction model. These genes are closely linked to the onset and progression of AD and can serve as reliable biomarkers. Furthermore, AD was categorized into three subtypes through consensus clustering analysis and the differences in gene expression, clinical features, immune infiltration, and pathway enrichment were examined. Furthermore, potential drugs for the treatment of each AD subtype were identified. The findings observed in the present study can help to deepen the understanding of the underlying disease mechanisms of AD and enable the development of precision medicine and personalized treatment approaches.The NCBI Gene Expression Omnibus (GEO) public database (https://www.ncbi.nlm.nih.gov/geo/) was used to search for gene expression data, and these data were obtained using the "GEOquery" R package. The data comprised two datasets, GSE109887 (GPL10904 platform) and GSE132903 (GPL10558 platform), which were merged to form a total of 130 and 143 samples from healthy individuals and patients with AD, respectively. The samples were collected from the MTG, a site of early AD pathology (18,19). Merged data were processed to eliminate batch effects from different platforms and to normalize the data using the "sva" package. A principal component analysis (PCA) was subsequently performed to assess data combinations. The performance of the prediction model was validated using the GSE5281 dataset (GPL570 platform), which contained 74 and 87 samples from healthy individuals and patients with AD, respectively, and the samples were obtained from several brain regions including the entorhinal cortex, hippocampus, medial temporal gyrus, posterior cingulate, superior frontal gyrus, and primary visual cortex.