Identification of moyamoya disease-related mitochondrial genes through machine learning based on a comprehensive gene expression database
10.3760/cma.j.issn.1673-4165.2025.01.004
- VernacularTitle:基于基因表达综合数据库通过机器学习鉴定烟雾病相关线粒体基因
- Author:
Yanbing HU
1
;
Xiong GUO
;
Adili ROUZI·
;
Hongbo PAN
Author Information
1. 喀什地区第二人民医院神经外科，喀什 844000
- Keywords:
Moyamoya disease;
Mitochondria;
Databases, genetic;
High-throughput nucleotide sequencing;
Machine learning
- From:
International Journal of Cerebrovascular Diseases
2025;33(1):18-29
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To investigate the moyamoya disease (MMD) -related mitochondrial genes and to elucidate potential regulatory mechanisms.Methods:The GSE157628 and GSE189993 datasets of MMD from the Gene Expression Omnibus (GEO) database were downloaded as the training set and validation set, respectively. Mitochondrial-related gene sets were obtained from the MitoCarta 3.0 database. Weighted gene co-expression network analysis was used to screen the pathogenic module genes of MMD. Venn analysis was used to identify differentially expressed genes related to mitochondria. Functional enrichment analysis was used to elucidate potential molecular functions. Mitochondrial genes were screened using three machine learning methods, receiver operator characteristic (ROC) curve, and validation set expression analysis. A diagnostic nomogram model of MMD related features was developed based on diagnostic genes. The correlation between the identified genes and MMD immune infiltration characteristics, the biological pathways involved, and potential therapeutic drugs with interactions were analyzed.Results:A total of 15 differentially expressed mitochondria-related genes were identified based on GSE157628 dataset, involving the processes related to small molecule metabolism and mitochondrial function. Three machine learning algorithms, ROC curves and expression analysis finally identified two mitochondria-related diagnostic genes: NUDT8 and RDH13. The diagnostic nomogram model developed based on these two genes had good diagnostic ability for MMD. Immune infiltration analysis showed that there was significant difference in plasma cells between the MMD group and the control group. In addition, the two diagnostic genes interacted with several therapeutic drugs. Conclusion:NUDT8 and RDH13 are identified as potential mitochondrial-related genes of MMD, and they may serve as therapeutic targets for progressive vascular occlusion in MMD.
