Related Experiment Video
Updated: Jul 30, 2025
A Machine Learning Approach to Design an Efficient Selective Screening of Mild Cognitive Impairment
Published on: January 11, 2020
Identification and validation of biomarkers based on cellular senescence in mild cognitive impairment
Songmei Ma1,2, Tong Xia1, Xinyi Wang1,3,4,5
1Department of Anesthesiology, The Third Central Clinical College of Tianjin Medical University, Tianjin, China.
Background:
Mild cognitive impairment (MCI), a syndrome defined as decline of cognitive function greater than expected for an individual's age and education level, occurs in up to 22.7% of elderly patients in United States, causing the heavy psychological and economic burdens to families and society. Cellular senescence (CS) is a stress response that accompanies permanent cell-cycle arrest, which has been reported to be a fundamental pathological mechanism of many age-related diseases. This study aims to explore biomarkers and potential therapeutic targets in MCI based on CS.
Methods:
The mRNA expression profiles of peripheral blood samples from patients in MCI and non-MCI group were download from gene expression omnibus (GEO) database (GSE63060 for training and GSE18309 for external validation), CS-related genes were obtained from CellAge database. Weighted gene co-expression network analysis (WGCNA) was conducted to discover the key relationships behind the co-expression modules. The differentially expressed CS-related genes would be obtained through overlapping among the above datasets. Then, pathway and GO enrichment analyses were performed to further elucidate the mechanism of MCI. The protein-protein interaction network was used to extract hub genes and the logistic regression was performed to distinguish the MCI patients from controls. The hub gene-drug network, hub gene-miRNA network as well as transcription factor-gene regulatory network were used to analyze potential therapeutic targets for MCI.
Results:
Eight CS-related genes were identified as key gene signatures in MCI group, which were mainly enriched in the regulation of response to DNA damage stimulus, Sin3 complex and transcription corepressor activity. The receiver operating characteristic curves of logistic regression diagnostic model were constructed and presented great diagnostic value in both training and validation set.
Conclusion:
Eight CS-related hub genes - SMARCA4, GAPDH, SMARCB1, RUNX1, SRC, TRIM28, TXN, and PRPF19 - serve as candidate biomarkers for MCI and display the excellent diagnostic value. Furthermore, we also provide a theoretical basis for targeted therapy against MCI through the above hub genes.
