知镜 Cognispect · 医学文献与研究工作台

Integrated transcriptomic analysis and machine learning identify AURKA and CENPM as candidate biomarkers in focal segmental glomerulosclerosis.

Lin X, Huang G, Sun Q, Zhang X.
PloS one · 2026-11-10 · PMID 43007984

文献摘要

Background Focal segmental glomerulosclerosis (FSGS) is a leading cause of steroid-resistant nephrotic syndrome and progression to end-stage renal disease, but existing therapeutics and diagnostics are limited. Therefore, reliable diagnostic biomarkers that contribute to FSGS therapy and diagnostics are urgently needed. Methods Gene expression data for FSGS (GSE104948; 37 samples: 16 FSGS vs 21 controls) were obtained from GEO. A weighted gene co-expression network analysis (WGCNA) was employed. Key genes within significant modules were identified using Support Vector Machine Recursive Feature Elimination (SVM-RFE), random forest (RF), and cytoHubba Maximal Clique Centrality (MCC). Functional enrichment analyses were executed, while diagnostic accuracy was evaluated by receiver operating characteristic (ROC) analysis. Immune cell infiltration (ICI) levels were quantified using CIBERSORT, and validation was conducted with an independent external dataset (GSE129973; 20 FSGS vs 20 controls). Results The green-yellow module demonstrated robust association with the FSGS phenotype, from which 39 candidate genes were selected. Integration of RF, SVM-RFE and cytoHubba analyses pinpointed two primary hub genes: AURKA, which was recovered consistently across cross-validation folds, and CENPM, a less consistently recovered candidate. Functional enrichment underscored the involvement of these genes in pivotal biological processes, particularly mitosis and regulation of the cell cycle. Analysis of immune infiltration revealed significant shifts in the immune microenvironment (IME) of FSGS, correlating elevated expression of AURKA and CENPM with altered ICI patterns. ROC curve analyses showed the diagnostic performance of these genes, with AURKA achieving an AUC of 0.95 and CENPM 0.94 in the training dataset (independent validation: AURKA AUC 0.797, CENPM AUC 0.720). These results showed a consistent direction of change between FSGS patients and healthy controls in an independent validation cohort. Conclusion Using a combined analytical approach involving WGCNA, machine learning (ML) algorithms, and protein-protein interaction (PPI) networks, this study identified AURKA and CENPM as exploratory tissue-based candidate biomarkers for FSGS, with AURKA being the more robustly supported of the two; both require functional validation and independent confirmation before any therapeutic application can be considered.

在知镜精读原文来源DOI 原文

知镜提供医学文献检索、阅读笔记、人体图谱、统计绘图与研究复现工具。