Construction and evaluation of Alzheimer’s disease diagnostic prediction model based on genes involved in mitophagy

Frontiers in Aging Neuroscience · 5 authors, 2 centres

AI SUMMARY

FIDELITY 100%

POPULATIONPatients with Alzheimer's disease (AD) and healthy controls from GEO datasets (GSE63061, GSE122063, GSE63060, GSE5281), plus 10 AD patients and 10 healthy controls for clinical validation.

INTERVENTIONDiagnostic prediction model based on expression levels of four mitophagy-related genes (OPTN, PTGS2, TOMM20, VDAC1).

COMPARISONAD samples vs. control samples; model predictions vs. actual clinical diagnosis.

This summary was generated by AI from a single paper. It has not been reviewed by a clinician and is not clinical advice. Verify against the source before acting on it.

AGENT 01 · STRUCTURED SUMMARY

TRACE TO SOURCE

“

TL;DR

This study identified four mitophagy-related genes (OPTN, PTGS2, TOMM20, and VDAC1) as potential diagnostic biomarkers for Alzheimer's disease (AD) using bioinformatics and machine learning. A logistic regression diagnostic model combining these genes achieved an AUC of 0.965 in the training dataset and 0.806 in an independent validation dataset. The findings suggest that peripheral blood expression of these genes could aid in early AD diagnosis, though protein-level and animal model validation are still needed.

Full summary

3,911 CHARS

**Background:** Alzheimer's disease (AD) is a common neurodegenerative disease with insidious onset, making early diagnosis difficult. Mitophagy, the selective autophagic removal of dysfunctional mitochondria, is crucial for neuronal health, and its dysfunction is implicated in AD. However, no study had systematically identified mitophagy-related diagnostic biomarkers for AD. This study aimed to screen mitophagy-related genes with diagnostic potential and construct a diagnostic prediction model.

**Methods:** The AD gene expression profile GSE63061 was downloaded from GEO. After normalization, protein-coding genes were extracted and compared with 137 mitophagy-related genes obtained from databases (Pathway Unification, GO, KEGG, MSigDB) and literature. Differentially expressed mitophagy-related genes (DE-MRGs) were identified using limma (|log2FC| >1, FDR <0.05, p<0.05). Functional enrichment (GO/KEGG), PPI network analysis (STRING, Cytoscape MCODE), and WGCNA (on GSE122063, 56 AD, 44 controls) were performed. Overlapping genes from PPI hub genes and WGCNA core modules were identified. Machine learning methods—univariate analysis, random forest, LASSO regression (λ=0.02, R²=0.80), and SVM classification—were used to select diagnostic markers. Multivariate logistic regression established the final diagnostic model. Model performance was evaluated by ROC curves (GSE122063 and GSE63061 as validation sets), nomogram, C-index, and calibration curves. Expression of key genes was validated in independent datasets (GSE63060, GSE5281), an AD cell model (SH-SY5Y + Aβ25-35, 25 μM), and peripheral blood from 10 AD patients and 10 healthy controls via RT-qPCR.

**Key Results:** From 114 overlapping mitophagy-related genes in GSE63061, 72 DE-MRGs were identified (63 upregulated, 9 downregulated). Top upregulated genes included UBC (logFC=3.90), UBB (logFC=3.58), TSPO (logFC=3.28); top downregulated included TOMM20 (logFC=-3.91), MAP1LC3B (logFC=-3.85), TOMM40 (logFC=-3.82). GO/KEGG enrichment showed involvement in autophagy, apoptosis, NF-κB signaling, ferroptosis, and neurological diseases. PPI analysis yielded 33 hub genes; WGCNA identified the turquoise module (576 genes) as most correlated with AD (r=-0.81, p=6e-08). Five overlapping genes (DNM1L, OPTN, PTGS2, TOMM20, VDAC1) were identified. Univariate analysis and random forest (accuracy 100% training, 73.3% test, F1=84.6%) selected OPTN, PTGS2, TOMM20, and VDAC1. LASSO and SVM (100% accuracy in both training and test sets, SVM-ROC AUC=100%) confirmed these four genes. Multivariate logistic regression yielded the diagnostic formula: y = 1.231 + 0.033×OPTN + 0.032×VDAC1 - 0.135×PTGS2 - 0.048×TOMM20. The model achieved AUC=0.965 in GSE122063 and AUC=0.806 in GSE63061. The nomogram (genes + age) had a C-index of 0.730, with good calibration. Individual gene AUCs in GSE63061 were: VDAC1=0.830, TOMM20=0.784, PTGS2=0.742, OPTN=0.726. GSEA showed these genes are involved in neurodegenerative disease pathways (PD, AD, HD, ALS), neural function, metabolism, and immune pathways. Validation in GSE5281 (brain tissue), GSE63060 (peripheral blood), AD cell model, and clinical blood samples confirmed consistent expression patterns: OPTN and VDAC1 upregulated, PTGS2 and TOMM20 downregulated in AD (all p<0.05).

**Clinical Implications:** This study provides a novel diagnostic prediction model based on four mitophagy-related genes (OPTN, PTGS2, TOMM20, VDAC1) that can be measured in peripheral blood, offering a non-invasive approach to aid early AD diagnosis. The consistent expression trends between brain tissue and peripheral blood support blood-based diagnostics. The model may help guide clinical decision-making and provide new therapeutic targets. However, limitations include lack of protein-level validation, absence of animal model confirmation, and the need for model updates as mitophagy gene databases expand.

PICO

PPOPULATION

Patients with Alzheimer's disease (AD) and healthy controls from GEO datasets (GSE63061, GSE122063, GSE63060, GSE5281), plus 10 AD patients and 10 healthy controls for clinical validation.

IINTERVENTION

Diagnostic prediction model based on expression levels of four mitophagy-related genes (OPTN, PTGS2, TOMM20, VDAC1).

OOUTCOME

Diagnostic accuracy measured by area under the receiver operating characteristic curve (AUC), C-index, calibration curves, and differential gene expression validation.

STUDY TYPE

other

SPECIALTY

neurology

SUMMARISED BY

AI pipeline

FIDELITY CHECK

100% · A

ORIGINAL PAPER

PMC10077494

Construction and evaluation of Alzheimer’s disease diagnostic prediction model based on genes involved in mitophagy

Frontiers in Aging Neuroscience

READ THE FULL PAPER

JUMP TO SECTION

ABSTRACT

Abstract Introduction Alzheimer’s disease (AD) is a common neurodegenerative disease. The concealment of the disease is the difficulty of its prevention and treatment. Previous studies have shown that mitophagy is crucial to the development of AD. However, there is a lack of research on the identification and clinical significance of mitophagy-related genes in AD. Therefore, the purpose of this study was to identify the mitophagy-related genes with the diagnostic potential for AD and establish a diagnostic model for AD. Methods Firstly, we download the AD gene expression profile from Gene Expression Omnibus (GEO). Limma, PPI, functional enrichment analysis and WGCNA were used to screen the differential expression of mitophagy-related AD gene. Then, machine learning methods (random forest, univariate analysis, support vector machine, LASSO regression and support vector machine classification) were used to identify diagnostic markers. Finally, the diagnostic model was established and evaluated by ROC, multiple regression analysis, nomogram, calibration curve and other methods. Moreover, multiple independent datasets, AD cell models and AD clinical samples were used to verify the expression level of characteristic genes in the diagnostic model. Results In total, 72 differentially expressed mitophagy-related related genes were identified, which were mainly involved in biological functions such as autophagy, apoptosis and neurological diseases. Four mitophagy-related genes (OPTN, PTGS2, TOMM20, and VDAC1) were identified as biomarkers. A diagnostic prediction model was constructed, and the reliability of the model was verified by receiver operating characteristic (ROC) curve analysis of GSE122063 and GSE63061. Then, we combine four mitophagy-related genes with age to establish a nomogram model. The ROC, C index and calibration curve show that the model has good prediction performance. Finally, multiple independent datasets, AD cell model samples and clinical peripheral blood samples confirmed that the expression levels of four mitophagy-related genes were consistent with the results of bioinformatics analysis. Discussion The analysis results and diagnostic model of this study are helpful for the follow-up clinical work and mechanism research of AD. Keywords: mitophagy, Alzheimer’s disease, biomarkers, diagnostic model, machine learning

CITATION EXPORT

Yongyan Pei, Sijia Chen, Fengling Zhou et al.. Construction and evaluation of Alzheimer’s disease diagnostic prediction model based on genes involved in mitophagy. Frontiers in Aging Neuroscience. (2023). https://doi.org/10.3389/fnagi.2023.1146660

AUTHORS

5 · 2 CENTRES

Yongyan Pei, Sijia Chen, Fengling Zhou, Tao Xie, Hua Cao

1School of Chemistry and Chemical Engineering, Guangdong Pharmaceutical University, Zhongshan, China
2Department of Neurology, The Third Affiliated Hospital of Naval Medical University, Shanghai, China