Table4_Identification of mitophagy-related genes with potential clinical utility in myocardial infarction at transcriptional level.docx

BackgroundMyocardial infarction (MI) ranks among the most prevalent cardiovascular diseases. Insufficient blood flow to the coronary arteries always leads to ischemic necrosis of the cardiac muscle. However, the mechanism of myocardial injury after MI remains unclear. This article aims to explore the potential common genes between mitophagy and MI and to construct a suitable prediction model.
MethodsTwo Gene Expression Omnibus (GEO) datasets (GSE62646 and GSE59867) were used to screen the differential expression genes in peripheral blood. SVM, RF, and LASSO algorithm were employed to find MI and mitophagy-related genes. Moreover, DT, KNN, RF, SVM and LR were conducted to build the binary models, and screened the best model to further external validation (GSE61144) and internal validation (10-fold cross validation and Bootstrap), respectively. The performance of various machine learning models was compared. In addition, immune cell infiltration correlation analysis was conducted with MCP-Counter and CIBERSORT.
ResultsWe finally identified ATG5, TOMM20, MFN2 transcriptionally differed between MI and stable coronary artery diseases. Both internal and external validation supported that these three genes could accurately predict MI withAUC = 0.914 and 0.930 by logistic regression, respectively. Additionally, functional analysis suggested that monocytes and neutrophils might be involved in mitochondrial autophagy after myocardial infarction.
ConclusionThe data showed that the transcritional levels of ATG5, TOMM20 and MFN2 in patients with MI were significantly different from the control group, which might be helpful to further accurately diagnose diseases and have potential application value in clinical practice.

背景：心肌梗死（Myocardial infarction, MI）是最常见的心血管疾病之一。冠状动脉血流灌注不足常导致心肌缺血性坏死，但心肌梗死后心肌损伤的具体机制尚未阐明。本研究旨在探讨线粒体自噬（mitophagy）与心肌梗死之间的潜在共通基因，并构建适宜的预测模型。

方法：本研究采用两项基因表达综合数据库（Gene Expression Omnibus, GEO）数据集（GSE62646与GSE59867）筛选外周血差异表达基因；采用支持向量机（Support Vector Machine, SVM）、随机森林（Random Forest, RF）与最小绝对收缩和选择算子（Least Absolute Shrinkage and Selection Operator, LASSO）算法筛选与心肌梗死及线粒体自噬相关的基因。此外，采用决策树（Decision Tree, DT）、K近邻（K-Nearest Neighbor, KNN）、随机森林、支持向量机与逻辑回归（Logistic Regression, LR）构建二分类模型，并筛选最优模型分别进行外部验证（数据集GSE61144）与内部验证（10折交叉验证及Bootstrap验证）；比较不同机器学习模型的预测性能。同时，采用MCP-Counter与CIBERSORT工具开展免疫细胞浸润相关性分析。

结果：本研究最终筛选出在心肌梗死与稳定型冠状动脉疾病患者中存在转录差异的ATG5、TOMM20及MFN2基因。内部与外部验证结果均显示，基于这三个基因构建的逻辑回归模型可准确预测心肌梗死，其曲线下面积（Area Under the Curve, AUC）分别为0.914与0.930。功能富集分析提示，单核细胞与中性粒细胞可能参与心肌梗死后的线粒体自噬过程。

结论：本研究数据显示，心肌梗死患者外周血中ATG5、TOMM20与MFN2的转录水平与对照组存在显著差异，该结果可为疾病的精准诊断提供参考，并有望在临床实践中展现潜在应用价值。