基于生物信息学与机器学习识别NAFLD 与肌少症的共享候选诊断生物标志物

Identification of Shared Candidate Diagnostic Biomarkers for NAFLD and Sarcopenia Based on Bioinformatics and Machine Learning

摘要: 目的：探讨非酒精性脂肪性肝病(NAFLD)与肌少症共享的跨组织分子特征，并筛选候选诊断生物标志物。方法：基于GEO数据库中的NAFLD肝组织数据集GSE135251和肌少症骨骼肌数据集GSE111016，采用差异表达分析、加权基因共表达网络分析(WGCNA)、逻辑回归和LASSO筛选共享候选分子，并结合模块富集分析与单样本通路活性评分进行功能解释。结果：共鉴定出11个在两种组织中一致下调的假基因注释转录本，并构建了可区分病例与对照的共病评分。进一步优先确定4个跨组织核心特征基因：ATP5F1EP2、NME2P1、RPL37P2和RPL39P38。该四基因特征在肝组织和骨骼肌中均具有一定判别能力，AUC分别为0.983和0.802，跨组织迁移分析AUC为0.715。富集分析提示，氧化磷酸化、核糖体/翻译及核苷酸相关过程在两种组织中均发生协同扰动，并伴随组织依赖性的自噬相关模式。结论：NAFLD与肌少症之间可能共享一种翻译–生物能量学应激程序，ATP5F1EP2、NME2P1、RPL37P2和RPL39P38可作为潜在候选诊断生物标志物，为后续机制研究提供线索。

Abstract: Objective: To investigate the shared cross-tissue molecular features of nonalcoholic fatty liver disease (NAFLD) and sarcopenia, and to identify candidate diagnostic biomarkers. Methods: Based on the NAFLD liver tissue dataset GSE135251 and the sarcopenic skeletal muscle dataset GSE111016 from the GEO database, differential expression analysis, weighted gene co-expression network analysis (WGCNA), logistic regression, and LASSO were used to identify shared candidate molecules, combined with module enrichment analysis and single-sample pathway activity scoring for functional interpretation. Results: A total of 11 pseudogene-annotated transcripts consistently downregulated in both tissues were identified, and a comorbidity score capable of distinguishing cases from controls was constructed. Four cross-tissue core feature genes were further prioritized: ATP5F1EP2, NME2P1, RPL37P2, and RPL39P38. This four-gene signature showed discriminative ability in both liver tissue and skeletal muscle, with AUCs of 0.983 and 0.802, respectively, while the cross-tissue transfer analysis yielded an AUC of 0.715. Enrichment analysis indicated coordinated perturbations of oxidative phosphorylation, ribosome/translation, and nucleotide-related processes in both tissues, accompanied by tissue-dependent autophagy-related patterns. Conclusion: NAFLD and sarcopenia may share a common translation-bioenergetic stress program. ATP5F1EP2, NME2P1, RPL37P2, and RPL39P38 may serve as potential candidate diagnostic biomarkers and provide clues for subsequent mechanistic studies.