PO.BCS01.15 · 生物信息与计算
去除双联体提升单细胞分辨率并揭示NSCLC中的恶性转录程序
Doublet removal enhances single-cell resolution and uncovers malignant transcriptional programs in NSCLC
作者与单位 Authors & Affiliations
摘要 Abstract
中文摘要
背景：非小细胞肺癌（NSCLC）占肺癌病例的85%以上，包括腺癌、鳞状细胞癌和大细胞癌等多种亚型。单细胞RNA测序（scRNA-seq）的进展实现了对NSCLC肿瘤微环境的高分辨率分析，揭示了此前未被识别的细胞异质性，并鉴定出潜在的治疗靶点，包括免疫检查点抑制剂（ICI）通路。然而，双联体（两个细胞被捕获于同一液滴中而人为合并的表达谱）会通过改变聚类和差异基因表达而扭曲下游分析，最终干扰生物学解读。
目的：本研究评估了一种引入双联体去除的改良scRNA-seq工作流程，以改善细胞类型鉴定和下游生物学洞察。
方法：从NCBI下载公共scRNA-seq数据集GSE198099，并同时采用常规流程和引入DoubletCatcher算法的改良流程进行重新分析。DoubletCatcher生成人工双联体，基于邻近关系计算双联体评分，并去除超过设定阈值的细胞。使用基于R的流程鉴定差异表达基因（DEGs）（FDR < 0.01；|log₂FC| > 0.2）。进行基因本体（Gene Ontology）富集分析以识别富集通路。生成UMAP图和热图用于可视化。
结果：常规分析产生了17个分辨度较差的聚类，而改良流程得到14个界定清晰的聚类，同时消除了免疫特征中的虚假重叠。清晰的细胞身份得以恢复，包括CD4⁺ T细胞、CD8⁺ T细胞、B细胞、浆母细胞、巨噬细胞、单核细胞、肥大细胞、内皮细胞、II型上皮细胞，以及由癌细胞和干样癌细胞组成的三个不同的癌细胞聚类。DEG分析揭示了肿瘤组织中癌细胞和干样癌细胞相较于癌旁组织存在显著的转录差异。肿瘤区癌细胞表现出去分化、缺氧驱动的代谢重编程、炎症和免疫逃逸信号、增殖和细胞周期激活以及上皮-间质转化（EMT）。相比之下，来自癌旁组织的癌细胞呈现更分化的状态，提示在手术干预前可能具有更强的治疗反应性。
结论：去除双联体显著提高了聚类准确性和生物学可解读性，揭示了不同的癌细胞、干样癌细胞和免疫细胞群体，并揭示了肿瘤区癌细胞中关键的活跃恶性程序。这一改良方法增强了单细胞分析的可靠性，并为NSCLC肿瘤生物学提供了进一步的洞见。
查看英文原文 English abstract
Background: Non-small cell lung cancer (NSCLC), accounting for more than 85% of lung cancer cases. includes several subtypes such as adenocarcinoma, squamous cell carcinoma, and large cell carcinoma. Advances in single-cell RNA sequencing (scRNA-seq) have enabled high-resolution profiling of the NSCLC tumor microenvironment, revealed previously unrecognized cellular heterogeneity, and identified potential therapeutic targets, including immune checkpoint inhibitor (ICI) pathways. However, doublets (artificially merged profiles of two cells captured in one droplet) can distort downstream analysis by altering clustering and differential gene expression, ultimately confounding biological interpretation.
Objective: This study evaluated a refined scRNA-seq workflow that incorporates doublet removal to improve cell-type identification and downstream biological insights.
Methods: Public scRNA-seq dataset GSE198099 was downloaded from NCBI and reanalyzed using both routine pipelines and a refined workflow that incorporates the DoubletCatcher algorithm. DoubletCatcher generates artificial doublets, computes doublet scores based on neighbor relationships, and removes cells exceeding the defined threshold. Differentially expressed genes (DEGs) were identified using R-based pipelines (FDR < 0.01; |log₂FC| > 0.2). Gene Ontology enrichment was performed to identify enriched pathways. UMAPs and heatmaps were generated for visualization.
Results: Routine analysis produced 17 poorly resolved clusters, while the refined workflow yielded 14 well-defined clusters, while eliminating spurious overlaps in immune signatures. Clear cell identities were recovered, including CD4⁺ T cells, CD8⁺ T cells, B cells, plasmablasts, macrophages, monocytes, mast cells, endothelial cells, type II epithelial cells, and three distinct cancer cell clusters comprising cancer cells and stem-like cancer cells. DEG analysis revealed significant transcriptional differences of cancer cells and stem-like cancer cells in tumor tissues compared with adjacent Tumor-region cancer cells exhibited de-differentiation, hypoxia-driven metabolic reprogramming, inflammatory and immune-evasive signaling, proliferation and cell-cycle activation, and epithelial-mesenchymal transition (EMT). In contrast, cancer cells from adjacent tissues showed more differentiated states, suggesting potentially greater therapeutic responsiveness before surgical intervention.
Conclusion: Removing doublets markedly improved cluster accuracy and biological interpretability, revealed distinct cancer cell, stem-like cancer cell, and immune cell populations, and uncovered key active malignant programs in tumor-region cancer cells. This refined method enhances the reliability of single-cell analysis and provides further insights into NSCLC tumor biology.
利益披露 Disclosure
B. Jin, None..
E. Liu, None..
A. Lei, None..
Q. Wang, None.