基于机器学习算法和生信分析挖掘NSCLC的脂质代谢相关诊断标志物及其免疫细胞浸润特征

Identified Lipid Metabolism-Related Biomarkers and Immune Cell Infiltration Characteristics in NSCLC Based on Machine Learning Algorithms and Bioinformatics Analysis

摘要: 目的：越来越多证据表明，非小细胞肺癌(NSCLC)的发病与脂质代谢异常有关，因此靶向代谢途径可能是一种有效的治疗策略。本研究旨在结合生物信息学和机器学习分析，以确定NSCLC的脂质代谢相关的诊断标志基因。方法：我们利用基因表达总库(GEO)中的基因表达数据集筛选出差异表达基因(DEGs)，并将其与加权基因共表达网络(WGCNA)筛选出的主要模块基因相结合，然后再与732个脂质代谢基因相交。通过最小绝对收缩和选择算法(LASSO)和随机森林算法(RF)确定诊断标志物。利用接收者操作特征曲线(ROC)确认了其有效性。此外，通过CIBERSORT算法研究诊断标志物与浸润免疫细胞之间的关联。最后，进行实验来验证我们的发现。结果：在关键模块基因、DEGs和脂质代谢基因取交集后，共发现了71个交叉基因。通过机器学习算法确定了DPEP2、ACADL、BDH2和CAVl为潜在的生物标志物，其ROC曲线下面积(AUC)值分别为0.999、0.999、0.996和0.997。验证集分析和qPCR结果与我们的发现一致。免疫细胞浸润分析表明，所有诊断特征都可能在不同程度上与NSCLC的多种免疫细胞相关。结论：我们利用机器学习和生物信息学方法确定了NSCLC中与脂质代谢相关的诊断特征基因，为靶向NSCLC异常脂质代谢的治疗提供了依据。

Abstract: Objectives: Targeting the metabolic pathways may be a potentially effective therapy strategy for non-small cell lung cancer (NSCLC), as increasing evidence associates the disease’s development to dysregulated lipid metabolism. Our research combines bioinfommatics and machine learning analysis to identify diagnostic hallmark genes for NSCLC. Methods: Gene expression datasets were available from the Gene Expression Omnibus (GO). First, differentially expressed genes (DEGs) were found and combined with the major modular genes selected by the weighted gene co-expression network (WGCNA) and then intersected with a set of 732 lipid metabolism genes. The biomarkers were identified by applying least absolute shrinkage and selection operator (LASSO) and random forest (RF) methods. The effectiveness and discrimination of the hub genes were confirmed using the receiver operating characteristic curve. Additionally, the association between diagnostic markers and infiltrating immune cells was investigated by calculating relative subsets of RNA transcripts (CIBERSORT). Finally, qRT-PCR experiments were conducted to validate our findings. Results: Overall, 71 lipid metabolism-related genes were detevted after overlapping key module genes and DEGs. We identified four key genes-DPEP2, ACADL, BDH2, and CAVl—as potential biomarkers, with area under the curve (AUC) values of 0.999, 0.999, 0.996, and 0.997, respectively, in the ROC curves. As indicated by the immune cell infiltration analysis, multiple immune cells may be involved in the development of NSCLC. Additionally, all diagnostic characteristics may correlate with immune cells to varying degrees. Conclusions: The hallmarks related to lipid metabolism (DPEP2, ACADL, BDH2 and CAV1) were identified using machine learning and bioinfommatics. Our study identifies potential diagnostic candidate genes for NSCLC and provides a basis for targeting aberrant lipid metabolism in NSCLC therapy.