Machine Learning and Experimental Validation of m6A RNA Methylation Related Signatures for Risk Prediction, Diagnostic Biomarkers, and Immune Subtypes in Chronic Kidney Disease.

BACKGROUND

N6-methyladenosine (m6A) RNA methylation, a pivotal epigenetic modification, has been implicated in the pathogenesis and progression of diverse diseases. This study sought to elucidate the functional contributions of m6A-related genes to the pathogenesis of chronic kidney disease (CKD) using a strategy that integrated machine learning and experimental validation, with the goal of identifying robust diagnostic biomarkers and novel molecular subtypes.

METHODS

Leveraging publicly available datasets, transcriptomic results of 53 patients with chronic kidney disease (CKD) as well as 8 healthy control individuals were collected. Differential expression analysis of m6A-related genes was performed, followed by the construction and comparison of random forest (RF) and support vector machine (SVM) models to predict CKD risk and identify diagnostic biomarkers. The key biomarkers were validated in the CKD model mice established by unilateral ureteral obstruction (UUO) using RT-qPCR and immunofluorescence analysis. Immune cell infiltration was assessed via ssGSEA analysis, and molecular subtypes were delineated through consensus clustering.

RESULTS

We identified 20 differentially expressed m6A-related genes in CKD. The RF model demonstrated superior performance in risk prediction and prioritized five key genes (CBLL1, ELAVL1, RBM15B, YTHDF1, METTL3) for constructing a diagnostic nomogram. Experimental validation confirmed the upregulation of CBLL1, ELAVL1, RBM15B, and YTHDF1, and the downregulation of METTL3 in CKD mice. Furthermore, we identified two distinct m6A-associated molecular subtypes (Clusters A and B) with divergent immune landscapes. Cluster B was characterized by a pro-inflammatory phenotype, featuring elevated Th17 cell infiltration and a reduced proportion of Th2 cells.

CONCLUSION

Beyond advancing the mechanistic understanding of m6A in CKD, this study provides a translatable risk prediction model and delineates distinct immune subtypes, offering valuable foundations for future clinical stratification, diagnostic refinement, and the development of personalized immunomodulatory therapies.