Uncovering the Impact of Aggrephagy in the Development of Alzheimer's Disease: Insights Into Diagnostic and Therapeutic Approaches from Machine Learning Analysis

Background:: Alzheimer's disease (AD) stands as a widespread neurodegenerative disorder marked by the gradual onset of memory impairment, predominantly impacting the elderly. With projections indicating a substantial surge in AD diagnoses, exceeding 13.8 million individuals by 2050, there arises an urgent imperative to discern novel biomarkers for AD. Methods:: To accomplish these objectives, we explored immune cell infiltration and the expression patterns of immune cells and immune function-related genes of AD patients. Furthermore, we utilized the consensus clustering method combined with aggrephagy-related genes (ARGs) for typing AD patients and categorized AD specimens into distinct clusters (C1, C2). A total of 272 candidate genes were meticulously identified through a combination of differential analysis and Weighted Gene Co-Expression Network Analysis (WGCNA). Subsequently, we applied three machine learning algorithms-namely random forest (RF), support vector machine (SVM), and generalized linear model (GLM)-to pinpoint a pathogenic signature comprising five genes associated with AD. To validate the predictive accuracy of these identified genes in discerning AD progression, we constructed nomograms. Results:: Our analyses uncovered that cluster C2 exhibits a higher immune expression than C1. Based on the ROC(0.956). We identified five characteristic genes (PFKFB4, PDK3, KIAA0319L, CEBPD, and PHC2T) associated with AD immune cells and function. The nomograms constructed on the basis of these five diagnostic genes demonstrated effectiveness. In the validation group, the ROC values were found to be 0.760 and 0.838, respectively. These results validate the robustness and reliability of the diagnostic model, affirming its potential for accurate identification of AD. Conclusion:: Our findings not only contribute to a deeper understanding of the molecular mechanisms underlying AD but also offer valuable insights for drug development and clinical analysis. The limitation of our study is the limited sample size, and although AD-related genes were identified and some of the mechanisms elucidated, further experiments are needed to elucidate the more in-depth mechanisms of these characterized genes in the disease. result: Our analyses revealed that in the machine learning training group (GSE33000 dataset), the RF, SVM, and GLM models achieved an area under the receiver operating characteristic curve (ROC AUC) values of 0.934, 0.956, and 0.728, respectively. Based on the ROC AUC, we selected the SVM model as our experimental method and identified five characteristic genes associated with AD. The validation group (GSE122063 and GSE109887 datasets) yielded ROC AUC values of 0.760 and 0.838, respectively.