A manganese metabolism-related gene signature stratifies prognosis and immunotherapy efficacy in kidney cancer

Abstract

Background: Manganese modulates tumorigenesis and immune regulation. High levels of manganese may promote cancer progression. While manganese toxicity causes renal tubular damage and chronic impairment, its association with kidney cancer remains poorly understood.

Methods: We systematically analyzed manganese metabolism genes in KIRC using the TCGA dataset. Through integrated bioinformatics approaches, including differential expression analysis, univariate Cox regression, and three machine learning algorithms (Boruta, GBM, and RFS), we identified prognosis-related MMCG. The Ward.D2 method was used to identify MMCG subtypes, while Lasso-cox regression analysis was performed to establish the MMCG risk model. The predictive performance was validated through time-dependent ROC analysis, calibration curves, and decision curve analysis.

Results: We identified 11 prognosis-related manganese metabolism core genes (MMCGs). KIRC patients were stratified into two clusters based on MMCG expression levels. Patients in Cluster I showed poorer outcomes, which were associated with tumour progression. The MMCG risk score was subsequently developed using LASSO-Cox regression analysis, and patients were classified into high- and low-risk groups. Survival analysis revealed that the outcomes of high-risk group patients were poorer than those of the low-risk group. Univariate and multivariate analyses confirmed the MMCG risk score as an independent prognostic biomarker. Pathway enrichment analysis showed differential enrichment of immune and metabolic pathways across subtypes and risk groups. We constructed a clinical nomogram incorporating the MMCG risk score and other clinical parameters, which demonstrated highly accurate predictive capabilities. Immune infiltration analysis and immune therapy response predictions indicated that patients in Cluster I and the high-risk group showed low responses to immune therapy.

Conclusion: Our findings provide a basis for clinical stratification strategies and future research on manganese-based interventions for renal cell carcinoma (RCC).

Keywords: Machine learning; Manganese metabolism; RCC; Risk signature; Subtype.

Fig. 1

The Flow Chart of this study

Fig. 2

Identification of Manganese Metabolism-Related Core Genes (MMCGs) A The volcano plot shows the differentially expressed manganese metabolism-related genes. B The heatmap represents expression levels of differentially expressed manganese metabolism-related genes across tumour and normal tissues. C T-SNE plot shows the distribution of tumour and normal samples on a two-dimensional coordinate axis based on manganese metabolism-related gene expression. D The forest plot shows the prognosis-related manganese metabolism genes; red indicates risk genes, and blue indicates protective genes. E–F Bubble plots highlight the enrichment of prognosis-related manganese metabolism genes in molecular functions (MF) and biological processes (BP). G Venn diagram shows the manganese metabolism core genes through three distinct machine learning algorithms

Fig. 3

Clinical relevance of manganese metabolism subtypes in the KIRC dataset A Heatmap shows expression levels of MMCGs in Manganese Metabolism Clusters I and II. B The survival curve shows the clinical outcomes of the patients in Manganese Metabolism Clusters I and II. C–F Correlations between manganese metabolism subtypes and clinical information (age, gender, grade, and stage). G–I Bubble plots show the gene ontology (BP, biological process; MF, molecular function; CC, cell component) enrichment of differentially expressed genes between Manganese Metabolism Clusters I and II. J Barplot depicted GSEA hallmark pathway enrichment between Cluster I and Cluster II. K Mutation characterization of KIRC patients across MMCG Clusters I and II

Fig. 4

Construction of MMCGs Risk Model A, B Lasso-Cox regression analysis process. C The bar plot represents the lasso regression coefficients of MMCG genes. D–F Survival curves demonstrate the differences in clinical outcomes between high and low MMCG risk groups across three independent RCC datasets. G–I ROC curves assess the predictive accuracy of the MMCG risk score for patient survival at 3, 5, and 7 years. J–L Multivariate Cox regression analysis, adjusting for age, gender, stage, grade, and MMCGs to determine the independent prognostic factors

Fig. 5

Heatmap shows the distribution differences of MMCG expression, patient age, gender, stage, grade, and MMCG subtypes in high and low MMCG risk groups across three independent RCC datasets

Fig. 6

Clinical Relevance, Nomogram Model, and Molecular Mechanisms A Nomogram predicts the survival rates of KIRC patients at 3, 5, and 7 years. B, C ROC curves and calibration curves evaluate the predictive accuracy of the nomogram in predicting a 3-/5-/7-year survival rate. D Survival curves show the prognostic differences between high-risk and low-risk groups based on the nomogram. E–G Bubble plots show the gene ontology enrichment (GO, BP, and CC) of differentially expressed genes between high-risk and low-risk groups. H Barplot shows GSEA hallmark pathway enrichment between high-risk and low-risk groups

Fig. 7

Immune Infiltration and Immunotherapy Response across MMCG Risk Groups. A, B Immune infiltration levels of innate and adaptive immune cells between MMCG risk groups. C Differences in the proportions of 23 immune cell types within MMCG risk groups. D Expression differences of seven immune checkpoints between high-risk and low-risk patients. E, F Tables compare the distribution of immune-responsive and non-responsive patients between high-/low-risk groups and Clusters I/II. G, H Boxplot compares the TIDE scores in high-risk and low-risk groups and Cluster I/II patients