Tian Jun, Zhang Lei, Shi Kexin, Yang Li
Department of Dermatology, Shaanxi Provincial People's Hospital, Xi'an, 710068, China.
Data Center of Shaanxi Provincial People's Hospital, Xi'an, 710068, China.
Immunol Res. 2025 Jan 11;73(1):30. doi: 10.1007/s12026-025-09593-x.
Mitophagy, the selective degradation of mitochondria by autophagy, plays a crucial role in cancer progression and therapy response. This study aims to elucidate the role of mitophagy-related genes (MRGs) in cutaneous melanoma (CM) through single-cell RNA sequencing (scRNA-seq) and machine learning approaches, ultimately developing a predictive model for patient prognosis. The scRNA-seq data, bulk transcriptomic data, and clinical data of CM were obtained from publicly available databases. The single-sample gene set enrichment analysis (ssGSEA) and weighted gene co-expression network analysis (WGCNA) were used to identify gene modules associated with mitophagy phenotypes. A machine learning framework employing ten different algorithms was used to develop the prognostic model. Based on scRNA-seq data, we identified 16 distinct cell subpopulations in melanoma, and melanoma cells exhibited significantly higher mitophagy scores. The turquoise module identified via WGCNA showed the strongest correlation with mitophagy scores. A prognostic model incorporating seven genes was developed through machine learning algorithms, achieving an average C-index of 0.754 across training and validation cohorts. Functionally, low-risk patients were enriched in interferon-gamma response and inflammatory processes, whereas high-risk patients showed enrichment in glycolysis regulation and signaling pathways such as KRAS and Wnt/β-catenin. Notably, low-risk patients demonstrated enhanced immune infiltration and greater sensitivity to immunotherapy. RT-qPCR validated the expression level of 7 model genes in human melanoma cell lines and normal melanocyte cell lines. Our study provides a comprehensive understanding of MRGs in melanoma and presents a novel prognostic model. These findings enhance our understanding of the tumor microenvironment and may guide personalized treatment strategies for CM patients.
线粒体自噬，即通过自噬对线粒体进行选择性降解，在癌症进展和治疗反应中起着至关重要的作用。本研究旨在通过单细胞RNA测序（scRNA-seq）和机器学习方法阐明线粒体自噬相关基因（MRGs）在皮肤黑色素瘤（CM）中的作用，最终建立一个用于预测患者预后的模型。CM的scRNA-seq数据、批量转录组数据和临床数据均来自公开可用的数据库。采用单样本基因集富集分析（ssGSEA）和加权基因共表达网络分析（WGCNA）来识别与线粒体自噬表型相关的基因模块。使用一个采用十种不同算法的机器学习框架来开发预后模型。基于scRNA-seq数据，我们在黑色素瘤中鉴定出16个不同的细胞亚群，并且黑色素瘤细胞表现出显著更高的线粒体自噬评分。通过WGCNA鉴定出的绿松石模块与线粒体自噬评分显示出最强的相关性。通过机器学习算法建立了一个包含七个基因的预后模型，在训练和验证队列中的平均C指数达到0.754。在功能上，低风险患者在干扰素-γ反应和炎症过程中富集，而高风险患者在糖酵解调节以及KRAS和Wnt/β-连环蛋白等信号通路中富集。值得注意的是，低风险患者表现出增强的免疫浸润以及对免疫治疗的更高敏感性。RT-qPCR验证了7个模型基因在人黑色素瘤细胞系和正常黑素细胞系中的表达水平。我们的研究提供了对黑色素瘤中MRGs的全面理解，并提出了一种新的预后模型。这些发现加深了我们对肿瘤微环境的理解，并可能为CM患者指导个性化治疗策略。
