探讨结直肠癌肺转移相关关键基因及潜在分子机制，为结直肠癌肺转移早期诊断及靶向治疗提供参考。

从基因表达综合数据库(gene expression omnibus, GEO)获取GSE41258数据集，对结直肠癌原发样本与肺转移样本进行差异表达分析，筛选差异表达基因(differentially expressed genes, DEGs)。采用基因本体(gene ontology, GO)和京都基因与基因组百科全书(Kyoto encyclopedia of genes and genomes，KEGG)富集分析探讨生物学功能，通过蛋白质-蛋白质相互作用(protein-protein interaction, PPI)网络筛选候选基因。采用随机森林(random forest, RF)、逻辑回归(logistic regression, LR)和轻量级梯度提升机(light gradient boosting machine, LightGBM)算法联合夏普利加性解释(Shapley additive explanations, SHAP)分析筛选核心基因，结合人类蛋白质图谱(human protein atlas, HPA)数据库及基因集富集分析(gene set enrichment analysis, GSEA)进行验证和功能分析。

共筛选获得196个结直肠癌肺转移相关DEGs。GO及KEGG富集分析显示，肺转移相关基因主要富集于细胞外基质(extracellular matrix, ECM)组织重塑、细胞黏附调控及能量代谢等与肿瘤转移密切相关的生物学过程和信号通路，提示可能在结直肠癌细胞向肺组织迁移、定植及转移微环境重塑过程中发挥作用。RF、LR及LightGBM模型在测试集中曲线下面积(area under the curve, AUC)分别为0.830、0.852和0.942，LightGBM模型预测性能最佳。结合SHAP特征重要性分析最终确定ATP柠檬酸裂解酶(ATP citrate lyase, ACLY)和聚集蛋白聚糖(aggrecan, ACAN)为结直肠癌肺转移相关核心基因。HPA数据库验证结果显示，ACLY和ACAN蛋白在结直肠癌组织中表达水平高于正常组织，在结直肠癌进展过程中发挥重要作用。GSEA结果显示，ACAN高表达组主要富集于ECM受体相互作用及整合素相关通路；ACLY高表达组主要富集于补体与凝血级联反应、核糖体生成及剪接体等相关通路，可通过调控肿瘤代谢重编程及细胞外基质重塑参与结直肠癌肺转移过程。

基于GEO数据库转录组数据，通过差异表达分析、功能富集分析、PPI网络分析及机器学习筛选，识别出ACLY和ACAN两个与结直肠癌肺转移相关候选基因。ACLY与ACAN表达特征可能反映结直肠癌肺转移分子表型，有望发展为肺转移组织学诊断辅助标志物，或作为靶向干预候选分子。

To identify key genes associated with colorectal cancer lung metastasis and explore their potential molecular mechanisms, thereby providing references for the early diagnosis and targeted therapy of colorectal cancer lung metastasis.

The GSE41258 dataset was obtained from the Gene Expression Omnibus (GEO) database. Differential expression analysis was performed between primary colorectal cancer samples and lung metastatic samples to identify differentially expressed genes (DEGs). Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were conducted to investigate the biological functions of DEGs, and candidate genes were screened using a Protein-Protein Interaction (PPI) network. Subsequently, Random Forest (RF), Logistic Regression (LR), and Light Gradient Boosting Machine (LightGBM) algorithms combined with Shapley Additive Explanations (SHAP) analysis were employed to identify key genes. The Human Protein Atlas (HPA) database and Gene Set Enrichment Analysis (GSEA) were further used for validation and functional characterization.

A total of 196 differentially expressed genes (DEGs) associated with colorectal cancer lung metastasis were identified. GO and KEGG enrichment analyses revealed that these genes were mainly enriched in biological processes and signaling pathways related to extracellular matrix (ECM) remodeling, cell adhesion regulation, and energy metabolism, suggesting their potential roles in colorectal cancer lung metastasis and metastatic microenvironment remodeling. The area under the curve (AUC) values of the RF, LR, and LightGBM models in the testing set were 0.830, 0.852, and 0.942, respectively, with LightGBM showing the best predictive performance. Integrated machine learning and SHAP analyses identified ATP citrate lyase (ACLY) and aggrecan (ACAN) as the key genes associated with colorectal cancer lung metastasis. Validation using the HPA database demonstrated that ACLY and ACAN proteins were expressed at higher levels in colorectal cancer tissues than in normal tissues, further supporting their potential roles in colorectal cancer progression. GSEA revealed that the ACAN high-expression group was mainly enriched in extracellular matrix (ECM) receptor interaction and integrin-related pathways, whereas the ACLY high-expression group was primarily enriched in complement and coagulation cascades, ribosome biogenesis, and spliceosome-related pathways, suggesting potential roles for these genes in colorectal cancer lung metastasis through the regulation of tumor metabolic reprogramming and extracellular matrix remodeling.

In summary, based on transcriptomic data from the GEO database, this study identified ACLY and ACAN as candidate genes associated with colorectal cancer lung metastasis through differential expression analysis, functional enrichment analysis, PPI network analysis, and machine learning screening. The expression features of ACLY and ACAN may reflect the molecular phenotype of colorectal cancer lung metastasis and may potentially serve as auxiliary markers for histological diagnosis of lung metastasis or as candidate molecules for targeted intervention.