Spatial transcriptomics Chowdhury et al

Data is spatial transcriptomics from Pancreatic Cancer PDX tissue. The mice were treated with Control or fibrinogen ASOs and tumors were collected at week 4 post-implant. Raw FASTQ files and accompanying H&E images were processed with SpaceRanger (10x Genomics). Reads were aligned to a dual-species reference (GRCh37 + mm10), after which unique molecular indices (UMIs) were assigned to individual Visium spots to generate spot-level gene-expression matrices. Because a matched single-cell reference was unavailable, initial compartment identification used the unsupervised spatial deconvolution tool stDeconvolve. For each inferred “topic,” we inspected the top-ranked genes, classified them as human- or mouse-specific, and labelled the corresponding spots as cancer (human) or stromal (mouse). Topic-wise proportions were then aggregated to yield the total cancer and stromal fractions per slide.本数据集包含胰腺癌患者来源异种移植（Patient-Derived Xenograft, PDX）组织的空间转录组学数据。实验小鼠分为对照组与纤维蛋白原反义寡核苷酸（Antisense Oligonucleotides, ASOs）处理组，于移植后第4周收集肿瘤组织。 原始FASTQ文件及配套的苏木精-伊红（Hematoxylin-Eosin, H&E）染色图像通过SpaceRanger（10x Genomics平台）进行预处理。测序reads比对至双物种参考基因组（GRCh37 + mm10），随后将唯一分子标识符（Unique Molecular Indices, UMIs）分配至单个Visium芯片斑点，从而生成斑点水平的基因表达矩阵。 由于缺乏匹配的单细胞参考数据集，初始组织分区识别采用了无监督空间反卷积工具stDeconvolve。针对每个推断得到的"主题"，我们对排名靠前的基因进行分析，将其分类为人类特异性或小鼠特异性基因，并将对应斑点标记为癌（人类来源）或间质（小鼠来源）。随后对各主题的占比进行汇总，即可得到每张切片的总癌组织占比与间质组织占比。