整合多套人肺癌单细胞转录组公共数据，系统解析肺腺癌肿瘤细胞恶性增殖转录调控网络的分子机制，筛选驱动肿瘤增殖的核心基因与转录因子，为肺癌早期诊断和靶向治疗提供理论依据。

整合GSE131907、GSE149655、GSE148071及E-MTAB-6149四套肺癌scRNA-seq数据集，经质控、标准化及Harmony批次校正后获得199 489个高质量细胞。采用Louvain聚类和经典标记基因注释细胞类型，通过inferCNV结合高斯混合模型(gaussian mixture model, GMM)基于拷贝数变异(copy number variation, CNV)特征精准区分肿瘤上皮细胞与正常上皮细胞。运用Wilcoxon秩和检验筛选差异表达基因(differentially expressed genes, DEGs)。通过加权基因共表达网络分析(weighted gene co-expression network analysis，WGCNA)鉴定关键功能模块，利用GENIE3算法构建转录因子调控网络。在A549细胞中通过siRNA敲低高迁移率族蛋白B2(high mobility group box 2, HMGB2)，采用qRT-PCR验证核心转录因子及下游靶基因的表达变化。

整合数据集经批次校正后细胞均匀混合，共鉴定出20个细胞亚群，分为免疫细胞(65.18%)、上皮细胞(27.43%)、内皮细胞(4.60%)和成纤维细胞(2.79%)。基于CNV的GMM方法成功识别23 505个肿瘤细胞(占上皮细胞44.90%)，高置信度肿瘤细胞占80.59%。WGCNA鉴定出蓝色模块为与细胞周期高度相关的关键模块，其活性在S期和G2M期显著升高，核心基因均为经典细胞周期驱动基因。转录调控网络分析揭示FOXM1、E2F7、HMGB2等7个核心转录因子形成紧密互作网络。细胞实验证实，敲低HMGB2可显著下调FOXM1、E2F7、TOP2A等关键基因的表达。

整合多套肺癌单细胞转录组数据，构建肺癌上皮细胞异质性的单细胞图谱，阐明以HMGB2-FOXM1/E2F7/HMGA1-细胞周期基因为核心轴的转录调控网络驱动肺癌上皮细胞恶性增殖的分子机制。结果为肺癌的早期诊断标志物筛选和靶向治疗策略开发提供了新的理论依据和潜在分子靶点。

To systematically dissect the molecular mechanisms underlying the transcriptional regulatory network governing malignant proliferation of lung adenocarcinoma cells through integrated analysis of multiple public human lung cancer single-cell transcriptomic datasets, and to identify core genes and transcription factors driving tumor proliferation, thereby providing a theoretical basis for early diagnosis and targeted therapy of lung cancer.

Four public lung cancer scRNA-seq datasets (GSE131907, GSE149655, GSE148071, and E-MTAB-6149) were integrated and processed. After quality control, normalization, and batch correction using Harmony, a total of 199 489 high-quality cells were retained. Cell clustering was performed using the Louvain algorithm, and cell types were annotated using canonical marker genes. InferCNV combined with Gaussian mixture modeling (GMM) was applied to accurately distinguish tumor epithelial cells from normal epithelial cells based on copy number variation (CNV) profiles. Differentially expressed genes (DEGs) were identified using the Wilcoxon rank-sum test. Weighted gene co-expression network analysis (WGCNA) was performed to identify key functional modules, and the GENIE3 algorithm was used to construct transcription factor regulatory networks. Finally, HMGB2 was knocked down in A549 cells via siRNA transfection, and the expression changes of core transcription factors and their downstream target genes were validated by qRT-PCR.

After batch correction, cells from different datasets were uniformly mixed. A total of 20 cell clusters were identified and classified into four major lineages: immune cells (65.18%), epithelial cells (27.43%), endothelial cells (4.60%), and fibroblasts (2.79%). Using CNV-based GMM analysis, 23 505 tumor cells were successfully identified (44.90% of all epithelial cells), among which 80.59% were classified as high-confidence malignant cells. WGCNA identified the blue module as the key module highly associated with the cell cycle, whose activity was markedly elevated in S and G2/M phases, with all hub genes being canonical cell cycle drivers. Transcriptional regulatory network analysis uncovered a tightly interacting regulatory circuit consisting of seven core transcription factors, including FOXM1, E2F7, and HMGB2. Functional cell experiments confirmed that knockdown of HMGB2 significantly downregulated the expression of key genes including FOXM1, E2F7, and TOP2A.

This study successfully integrated multiple lung cancer single-cell transcriptomic datasets and established a comprehensive single-cell atlas of lung cancer epithelial cell heterogeneity. We elucidated the molecular mechanism by which the core transcriptional regulatory axis centered on HMGB2-FOXM1/E2F7/HMGA1-cell cycle genes drives malignant proliferation of lung cancer epithelial cells. These findings provide a novel theoretical basis and potential molecular targets for the identification of early diagnostic biomarkers and the development of targeted therapeutic strategies for lung cancer.