Literature-derived serum miRNA signatures associated with cognitive decline in Alzheimer’s disease: integrated analysis and machine learning-based diagnostic modeling

Because the clinical diagnosis of Alzheimer’s disease (AD) still relies largely on cognitive decline and A/T/N biomarkers remain costly and invasive, we aimed to identify literature-derived serum microRNAs (miRNAs) associated with cognitive function in AD and to re-evaluate them in a large public cohort (GSE120584) as potential adjunctive diagnostic biomarkers. We systematically searched Chinese- and English-language databases for studies reporting serum miRNA expression in patients with AD and its association with cognitive scale scores. Functional target genes were retrieved from miRTarBase, protein–protein interaction (PPI) networks were constructed using STRING, and functional modules were identified with the Cytoscape plug-in MCODE. Enrichment analyses were then performed for Gene Ontology (GO), KEGG, Reactome, and Hallmark gene sets. Differential expression analysis of GSE120584 was conducted using limma, and partial correlations between miRNA expression and age, sex, and apolipoprotein E ε4 (APOE ε4) allele count were calculated. Based on the correlation structure and nested cross-validation, optimal miRNA combinations were selected, and multiple diagnostic models were developed using age and sex as baseline clinical predictors. Model performance was evaluated using receiver operating characteristic (ROC) and precision-recall (PR) curves, calibration curves, and decision curve analysis (DCA). Twenty-three publications including 2,458 patients with AD and 2,139 controls were ultimately included. Twenty-two differentially expressed serum miRNAs were identified, of which 15 were positively and 7 were negatively correlated with Mini-Mental State Examination (MMSE) scores. Target genes of positively correlated miRNAs were enriched in PI3K/AKT/mTOR, Wnt, and TNF-α/NF-κB signaling pathways, whereas target genes of negatively correlated miRNAs were mainly involved in cell cycle regulation, the G2/M checkpoint, and oxidative stress responses. After matching literature-derived miRNAs with GSE120584 and expanding the candidate miRNA set, 25 significantly differentially expressed miRNAs were identified. Correlation-network analysis combined with nested cross-validation indicated that the minimal miRNA signatures with optimal diagnostic value were miR-211-5p alone (K1) and the three-miRNA panel miR-211-5p, miR-128–1-5p, and miR-128-3p (K3). When age and sex were added, the “K3 + age + sex” model showed the best performance (area under the curve AUC = 0.838, average precision AP = 0.934, Brier score = 0.162), yielding the highest sensitivity (0.563) and the best positive predictive value (PPV = 0.954) at specificity ≥ 0.90. By integrating published evidence with re-evaluation in a public serum cohort, we found that adding miR-211-5p and a miR-128-related panel to age- and sex-based models provided a modest but consistent improvement in diagnostic discrimination for AD. These miRNAs may serve as adjunctive peripheral blood biomarkers, although their clinical utility requires confirmation in independent external cohorts.