Identification of diagnostic biomarkers for sarcopenia based on bioinformatics and machine learning
10.3760/cma.j.cn115822-20231106-00055
- VernacularTitle:基于生物信息学和机器学习筛选肌少症的生物标志物
- Author:
Shijia WANG
1
;
Yu ZHANG
;
Jiayu GUO
;
Kang YU
Author Information
1. 苏州大学附属第二医院临床营养科 215004
- Keywords:
Bioinformatics;
Sarcopenia;
Machine learning;
Biomarkers
- From:
Chinese Journal of Clinical Nutrition
2023;31(6):321-329
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To identify characteristic genes in sarcopenia patients through bioinformatics and machine learning, and to explore the clinical relevance of characteristic genes in the diagnosis of sarcopenia.Methods:The microarray data of GSE25941, GSE38718 and GSE9103 associated with sarcopenia were downloaded from the GEO database, followed by identification of differentially expressed genes (DEGs) associated with sarcopenia. Subsequently, functional analysis of the DEGs was performed using gene ontology (GO) functional annotation and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis. The protein-protein interaction (PPI) network was constructed using STRING and Cytoscape, while biomarkers of sarcopenia were identified using LASSO regression and random forest analysis. The diagnostic performance of the characteristic gene was assessed by employing receiver operating characteristic (ROC) curve analysis. Furthermore, the expression levels of biomarkers for sarcopenia were validated using the external validation dataset of GSE28422. Finally, CIBERSORT was employed to analyze the infiltration of immune cells.Results:124 DEGs were identified between control and sarcopenia populations, which were primarily involved in growth factor receptor binding and cytokine activity. KEGG analysis revealed that the DEGs were predominantly associated with signaling pathways such as peroxisome proliferator-activated receptor signaling pathway, adipokine signaling pathway, Jak-STAT signaling pathway, and adenosine 5'-monophosphate (AMP)-activated protein kinase signaling pathway. Through machine learning techniques validated by ROC curve analysis and external datasets, three characteristic genes, namely DMRT2, FAM171A1, and ARHGAP36, were discovered. The infiltration analysis of immune cells revealed the potential involvement of mast cells, CD4 memory T cells, CD8 cells, γδT cells, and neutrophils in the pathophysiology of sarcopenia.Conclusion:DMRT2, FAM171A1 and ARHGAP36 can serve as diagnostic biomarkers of sarcopenia, and are closely related to the pathophysiological process of sarcopenia.