Single-cell dataset of renal cell carcinoma

We collected the single-cell transcriptome data of patients receiving ICB treatments, including ICB treatment-naïve tumor samples, ICB-resistant samples (post treatments) and ICB-sensitive samples for the investigation of ICB-resistant mechanisms. Besides, we also included adjacent normal tissues and PBMC samples for comparison. Taken together, we obtained a total of 6 adjacent normal, 9 ICB treatment-naïve tumor, 5 ICB-sensitive tumor, 4 ICB-resistant tumor and 2 PBMC samples.
We used the Cell Ranger software (Version 7.0.0) with default settings to align and quantify single-cell transcriptome data in FASTQ format published by 10x Genomics against the GRCh38 human reference genome[67]. The Cell Ranger software's quantified count matrix was loaded into the Seurat tool (Version 4.1.1) for further analyses.The cells were then subjected to quality control. Basically, cells with fewer than 500 identified genes and those with greater than 10% mitochondrial content were eliminated. To further exclude probable doublets, cells containing more than 8000 identified genes were discarded. Possible doublets predicted by the DoubletFinder software were eliminated so as not to impede the analyses. After filtering, samples containing fewer than 500 cells were deemed of poor quality and eliminated. More than 1000 thousand cells of good quality were preserved for further analysis. All individual high-quality single-cell samples were then curated into a single object and to eliminate batch effects. The dimensionality of this dataset was reduced through principal component analysis (PCA) with highly variable features, and the first 15 PCs were selected for investigation. Then, unsupervised clustering was approximated using the shared nearest-neighbor network produced by the Louvain algorithm and the edge weights between any two cells. Using the t-Distributed Stochastic Neighbor Embedding (t-SNE) and Uniform Manifold Approximation and Projection (UMAP) methods, the identified clusters were displayed. We identified the differentially expressed markers of the resultant clusters and used the default nonparametric test, the Wilcoxon rank sum test with Bonferroni correction, to label the cell clusters.

我们收集了接受免疫检查点阻断治疗（ICB）的患者的单细胞转录组数据，包括未接受ICB治疗的肿瘤样本、治疗后获得ICB耐药的样本以及ICB敏感样本，以探究ICB耐药机制。此外，我们还纳入了癌旁正常组织与外周血单个核细胞（PBMC）样本作为对照。总计收集得到6份癌旁正常组织样本、9份未接受ICB治疗的肿瘤样本、5份ICB敏感肿瘤样本、4份ICB耐药肿瘤样本以及2份PBMC样本。
我们使用Cell Ranger软件（版本7.0.0），以默认参数对10x Genomics发布的FASTQ格式单细胞转录组数据进行比对与定量，比对参考基因组为GRCh38人类参考基因组[67]。将Cell Ranger软件生成的定量计数矩阵加载至Seurat工具（版本4.1.1）中进行后续分析。
随后对细胞进行质量控制：具体而言，我们剔除了基因检出数少于500个、线粒体基因占比超过10%的细胞；为进一步排除潜在的双细胞（doublets），同时丢弃了基因检出数超过8000个的细胞。此外，我们还移除了由DoubletFinder软件预测得到的潜在双细胞，以避免对后续分析造成干扰。经过上述过滤后，细胞数少于500个的样本被判定为低质量样本并予以剔除。最终保留了超过100万个高质量细胞用于后续分析。
随后将所有高质量单细胞样本整合为单个对象，并消除批次效应。本数据集通过基于高可变特征的主成分分析（PCA）进行降维，并选取前15个主成分进行后续分析。随后基于Louvain算法生成的共享最近邻网络及任意两细胞间的边权重，通过无监督聚类对细胞进行分组。我们采用t分布邻域嵌入（t-SNE）与均匀流形近似与投影（UMAP）方法对鉴定得到的细胞簇进行可视化展示。我们鉴定了各细胞簇的差异表达标志物，并采用默认的非参数检验——带有Bonferroni校正的Wilcoxon秩和检验——对细胞簇进行注释。