Single-cell spatial transcriptomics (6k-plex CosMx SMI) dataset of human basal-cell carcinoma
DOI10.5281/zenodo.14330691Zenodo14330691MaRDI QIDQ6691346FDOQ6691346
Dataset published at Zenodo repository.
Massimo Andreatta, Santiago J. Carmona, Francois Kuonen, Laura Yerly
Publication date: 9 December 2024
Copyright license: Creative Commons Attribution 4.0 International
Single-cell spatial transcriptomics dataset of basal-cell carcinoma (BCC), collected using in situ analysis platform 6k-plex CosMx SMI (Bruker Nanostring). The dataset is provided as a Seurat object in .rds format, and consists of in situ transcripts for 6075 genes in 232,802 cells + metadata (see Metadata section below). Experimental conditions: The dataset is derived from tumor sections of four patients. For 2 patients (A and B), nodular-ulcerated BCC samples (as determined by histopathological reports) and of mixed morphology (i.e. consisting of nodular and infiltrative areas). For 2 patients (C and D), we collected samples in three different conditions: i) from baseline (first, diagnostic biopsy); ii) 1 week later from the previously biopsied, wounded site; iii) 1 week after initial biopsy from a distant, unwounded site in the same tumor. Data acquisition and annotation: Tumor sections from four different patients were arranged on 2 slides. For each slide, 45 Fields-of-views (FOVs) (0.51mm by 0.51mm) were selected based on the immunofluorescent staining with morphological markers (DAPI, PanCK, CD45, CD68). Tissue slides were subjected to in situ chemistry and imaging using the CosMx SMI instrument. CosMx scan data were uploaded to Nanostrings AtoMx spatial analysis platform, where pre-processing steps, imaging-barcode decoding and cell segmentation were performed. The pre-processed spatial transcriptomics data were exported to .rds files, and subsequent analyses were performed in R and Seurat. To predict cell type identities from CosMx in situ transcripts, we applied the InSituType algorithm, using a set of 7 confidently-annotated scRNA-seq samples (Yerly et al. 2022 cohort) as a reference for label transfer, to which we manually added an average expression profile for neutrophils derived from Zilionis et al. (2019). Cells with 100 detected transcripts and with 2% transcripts coming from negative control probes were labeled as Low quality cells. To define cancer cells that participate in homotypic or heterotypic interactions in our CosMx spatial datasets, we first calculated nearest neighbor graphs based on spatial coordinates using the BiocNeighbors package. We limited the neighbor search between cell centroids to a maximum distance D, corresponding to 2.5 times the average cell diameter in the CosMx dataset. If 90% of the cells within distance D of a given cancer cells were also cancer cells, the cancer cell was labeled as homotypic; otherwise it was labeled as heterotypic. Metadata: The dataset contains in situ transcripts for 6075 genes in 232,802 cells, as well as the following cell metadata: - pat_fov: identifies unique tissue slide and FOV combinations. Note: Run identifier "Run6057_Patient1" includes samples from Patient A and Patient C, run identifier "Run6057_Patient4" includes samples from Patient B and Patient D. See Patient_ID metadata column. - nCount_Nanostring: number of transcripts detected per cell - nFeature_Nanostring: number of unique genes detected per cell - cell Area, Aspect ratio, Height and Width - mean and max readouts for a panel of antibodies (PanCK, CD68, membrane staining, CD45 and DAPI) - Patient_ID, for four patients (A to D) - Area_ID: identifying 8 areas of BCC tissue - Condition: indicating the experimental condition of the sample - Ulcerated_area: wheher the FOV was annotated in an ulcerated BCC area - Dist_from_wound: for wounded samples, whether the FOV is located close or far from the wound - Condition2: concatenates Condition (see above) with distance from wound - Wound_direction: cardinal coordinates for the directionality of the wound with respect to the FOV - Morphology: whether the FOV was annotated as nodular or infiltrative in terms of HE morphology - celltype: predicted cell type annotation based on inSituType label transfer - celltype_prob: confidence score for cell type annotion - Meta-programs signature scores for MP1 to MP7, calculated with UCell - CAF signature scores (for wound-responding CAFs and baseline CAFs) - Invasiveness score: corresponds to MP7 minus MP2 scores - Invasive CAF score: corresponds to wound-responding CAF score minus baseline/unwounded CAF score - homotypic: for cancer cells, whether cell-cell interactions are homotypic (only with other cancer cells) or heterotypic (with stromal components). - X and Y coordinates of the centroid of each cell - neighbors vector: lists the IDs of the nearest neighbors of each cell - neighbors_encoding: the cell type distribution of the nearest neighbors; the vector is order according to the factor of cell types.
This page was built for dataset: Single-cell spatial transcriptomics (6k-plex CosMx SMI) dataset of human basal-cell carcinoma
