Data from: High-dimensional imaging of vestibular schwannoma reveals distinctive immunological networks across histomorphic niches in NF2-related schwannomatosis
Data files
Feb 03, 2025 version files 2.45 GB
-
cases.csv
1.23 KB
-
cell_level_expression.csv
362.38 MB
-
cell_level_metadata.csv
89.79 MB
-
images.zip
700.62 MB
-
masks.zip
14.56 MB
-
README.md
5.59 KB
-
VS_NF2_IMC_AnnData.h5ad
1.28 GB
Abstract
In this dataset, we have used imaging mass cytometry to identify and map the various populations in the human vestibular schwannomas tumor microenvironment (TME) using known markers for myeloid and schwann cells in vestibular schwannoma. This dataset consists of imaging mass cytometry data (16-bit TIFF images) for 17 vestibular schwannomas sourced from the Salford Royal NHS Trust Biobank. We identified various schwann, myeloid and T-cell subsets that has distinct spatial locations within the TME.
README: Imaging mass cytometry data from NF2-mutated vestibular schwannomas
https://doi.org/10.5061/dryad.wwpzgmstv
Description of the data and file structure
In this dataset, we have used imaging mass cytometry to identify and map the various populations in the human vestibular schwannomas tumor microenvironment (TME) using known markers for myeloid and schwann cells in vestibular schwannoma. This dataset consists of imaging mass cytometry data (16-bit TIFF images) for 17 vestibular schwannomas sourced from the Salford Royal NHS Trust Biobank. We identified various schwann, myeloid and T-cell subsets that has distinct spatial locations within the TME.
To generate this data, 5 um FFPE sections from 17 vestibular schwannomas were stained using the protocol recommended by Standard BioTools (https://www.standardbio.com/products/instruments/hyperion) using a panel of metal-conjugated antibodies. They were then imaged on the Hyperion using the standard settings. Regions of interest were identified on serial-cut H&E stained sections by a neuropathologist, annotating them Antoni A or Antoni B (see cases.csv). Raw TIFF images were then extracted from MCD files, and denoised using the IMC-Denoise method (https://www.nature.com/articles/s41467-023-37123-6). Single-cell information for each of the channels was extracted using the Bodenmiller pipeline (https://github.com/BodenmillerGroup/ImcSegmentationPipeline), creating a cell table detailing the raw mean single-cell expression of each of the markers in the panel, along with their X and Y locations in the region of interest. Individual markers were then normalised to the 99th percentile (ie, normalised from 0 to 1) for down stream analysis.
Files and variables
File: cases.csv
Description: Case metadata
Variables
- Case: Anonymised clinical case
- Disease_Severity: NF2 disease severity
- Treatment_Status: Avastin or untreated ('Naive')
- Tumour_Size(cm3): Tumour size calculated volumetrically via MRI
- Annual_Percentage_Growth: Annual growth rate adjusted for tumour size
File: cell_level_metadata.csv
Description: Cell-level metadata
Variables
- ROI: Region of interest
- X_loc: X location in image
- Y_loc: Y location in image
- Master_Index: Cell index over entire dataset
- Case: Anonymised clinical case
- Disease_Severity: NF2 disease severity
- Treatment_Status: Avastin or untreated
- Tumour_Size(cm3): Tumour size calculated volumetrically via MRI (* Note: not available for all cases*)
- Annual_Percentage_Growth: Annual growth rate adjusted for tumour size (* Note: not available for all cases*)
- Histopathology: Antoni A vs B (vestibular schwannoma pathology)
- population: Cell population, identified using leiden clustering
- hierarchy: High-level cell population/family, identified using leiden clustering
File: masks.zip
Description:
Segmentation masks created by Bodemiller pipeline (Note: Mask for region NF2VS10.1b was corrupted and lost).
File: cell_level_expression.csv
Description:
Processed (normalised to 99th expression within each marker) expression of markers in the panel
File: images.zip
Description:
Denoised images for all ROIs (Note: Images for region NF2VS10.1b were corrupted and lost)
File: VS_NF2_IMC_AnnData.h5ad
Description:
Processed data in AnnData format (https://anndata.readthedocs.io/en/stable/), where markers were normalised to 99th percentile, and were batch corrected using the BBKNN algorithm (https://github.com/Teichlab/bbknn), before UMAPs were created and populations identified using leiden clustering (https://www.nature.com/articles/s41598-019-41695-z).
Var (metadata):
- Indexed by name of antigen, no other information
Obs (cell metadata):
- ROI: Region of interest
- X_loc: X location in image
- Y_loc: Y location in image
- Master_Index: Cell index over entire dataset
- Case: Anonymised clinical case
- Disease_Severity: NF2 disease severity
- Treatment_Status: Avastin or untreated
- Tumour_Size(cm3): Tumour size calculated volumetrically via MRI (**Note:* not available for all cases*)
- Annual_Percentage_Growth: Annual growth rate adjusted for tumour size (**Note:* not available for all cases*)
- Histopathology: Antoni A vs B (vestibular schwannoma pathology)
- population: Cell population, identified using leiden clustering
- hierarchy: High-level cell population/family, identified using leiden clustering
Code/software
VS_NF2_IMC_AnnData.h5ad - AnnData object - https://anndata.readthedocs.io/en/stable/
Bodenmiller pipeline - https://github.com/BodenmillerGroup/ImcSegmentationPipeline
IMC_Denoise used for denoising - https://www.nature.com/articles/s41467-023-37123-6
Standard BioTools tissue staining protocol - https://www.standardbio.com/products/instruments/hyperion
Images - 16-bit TIFF images
Masks - TIFF segmentations masks where all pixels belonging to a cell are given the same value.
Methods
In brief, 5 um FFPE sections from 17 vestibular schwannomas (see cases.csv) were stained using the protocol recommended by Standard BioTools (https://www.standardbio.com/products/instruments/hyperion) using a panel of metal-conjugated antibodies. They were then imaged on the Hyperion using the standard settings. Regions of interest were identified on serial-cut H&E stained sections by a neuropathologist, labelling them Antoni A or Antoni B (see cases.csv). Raw TIFF images were then extracted from MCD files, and denoised using the IMC-Denoise method (https://www.nature.com/articles/s41467-023-37123-6). Denoised images are provided (images.zip). Single-cell information for each of the channels was extracted using the Bodenmiller pipeline (https://github.com/BodenmillerGroup/ImcSegmentationPipeline), with the resulting cell table (see cell_table.csv) detailing the raw mean single-cell expression of each of the markers in the panel, along with their X and Y locations in the region of interest. Meta-data for cells is also provided (cell_level_metadata.csv). Segmentation masks created by the Bodenmiller pipeline are also provided (masks.zip). Processed data is also provided as an AnnData object (VS_NF2_IMC_AnnData.h5ad), where markers were normalised to the 99th percentile, and population were identified by leiden clustering.