Abstract: BACKGROUND: Alzheimer's disease is a neurodegenerative disorder. Although amyloid-beta and tau proteins are core biomarkers for the diagnosis of Alzheimer's disease, exploring new biomarkers for disease diagnosis and treatment is still of great significance due to their heterogeneity and diagnostic limitations.
OBJECTIVE: To analyze the interplay between N6-methyladenosine epitranscriptomic modifications and ferroptosis-related genes in Alzheimer’s disease, and identify characteristic genes associated with Alzheimer’s disease pathogenesis through machine learning, bioinformatics analysis, and experimental validation, and to reveal their regulatory associations through the immune microenvironment, providing novel biomarkers for the early diagnosis and precise treatment of Alzheimer’s disease.
METHODS: The tissue transcriptomic data from GSE5281, GSE48350 (training sets) and GSE33000 (validation set) in the GEO database were integrated. N6-methyladenine regulatory factors with differential expression in Alzheimer’s disease from the training set were screened. The correlation between N6-methyladenine and ferroptosis genes was evaluated. Ferroptosis-related differential genes associated with N6-methyladenine were identified. Support vector machine recursive feature elimination algorithm combined with the Boruta feature selection model was used to determine the characteristic genes of Alzheimer’s disease. The functional modules of these characteristic genes were deciphered via gene set enrichment analysis. A logistic regression model integrated with the receiver operating characteristic curve were evaluated the diagnostic efficacy of the characteristic genes in the validation set. The single-sample gene set enrichment analysis was quantified immune cell infiltration levels and their regulatory associations of these cells with the characteristic genes were analyzed. Based on the ENCORI database, miRWalk 3.0 database, and NetworkAnalyst, the transcription factor/miRNA–mRNA regulatory network was constructed. Potential therapeutic compounds were further screened using the CTD database. The experimental validation in the hippocampal tissue of APP/PS1 double-transgenic mice was conducted using quantitative reverse transcription polymerase chain reaction and Western blotting.
RESULTS AND CONCLUSION: (1) Two significantly differentially expressed N6-methyladenosine regulatory factors were identified, namely Wilms’ tumor 1-associating protein (WTAP) and methyltransferase-like protein 14 (METTL14), along with a total of 16 ferroptosis-related genes associated with them. (2) Five hub characteristic genes were screened using machine learning, including fumarate hydratase, aspartate aminotransferase, GTPase HRas, metallothionein-3, histone-lysine n-methyltransferase SETD1B. (3) The functions of characteristic genes were enriched in the oxidative phosphorylation signaling pathway, Huntington's disease, Parkinson's disease, fatty acid degradation and metabolic pathway, and proteasome signaling pathway. (4) The area under the curve values of the constructed logistic regression diagnostic model in the training set and the validation set were 0.873 and 0.904, respectively, indicating excellent diagnostic efficiency of the feature genes. (5) Immune microenvironment analysis revealed that the GTPase HRas gene is significantly correlated with the levels of CC chemokine receptor family members and plasmacytoid dendritic cell infiltration. (6) The regulatory network containing 5 mRNA-37 miRNA-142 transcription factors was constructed, predicting 71 targeted therapeutic drugs. (7) Experimental verification showed that the mRNA and protein expression levels of aspartate aminotransferase, GTPase HRas and histone-lysine N-methyltransferase SETD1B in the hippocampus of APP/PS1 mice were significantly different (P < 0.05 or P < 0.01), which was consistent with the bioinformatics results. (8) These results reveal that fumarate hydratase, aspartate aminotransferase, GTPase HRas, metallothionein-3 and histone-lysine N-methyltransferase SETD1B as characteristic genes associated with the pathogenesis of Alzheimer's disease. Immune infiltration cell association analysis suggests that GTPase HRas may have the value of immunotherapy markers for Alzheimer's disease, providing a theoretical basis for early diagnosis and targeted therapy of the disease.

Key words: Alzheimer’s disease, N6-methyladenosine, ferroptosis, machine learning, immune infiltration, gene enrichment analysis, APP/PS1 transgenic mice, experimental validation