Spatial Tile

CODEX

For CODEX we accept image data in three different ways.

1. x, y coordinates

   The simplest representation consists of \(x\) and \(y\) coordinates corresponding to cell centers. This info should be provided in the anndata file under adata.obs['x'] and adata.obs['y']. The cells will be visualized as dots.

   See an example file.

2. Spatial tile

   A single channel image with pixel values representing cell membership can also be provided under adata.uns['spatial_idx']. The tile can have any shape (width, height). Pixel \((i, j)\) should be set to the index of a cell in adata (from 0 to the number of cells minus 1). A pixel value of -1 implies an empty location, and a value of -2 implies cell boundary (optional).

   Example file.

3. cytokit

   Lastly, we accept output produced by cytokit. You will need the following:

   • Segmented images

     These are .tif files named R001_X00\*_Y00\*.tif where the asterix character specifies the x and y coordinates of the tile (starting from 1). These images should have shape (z-planes, 4, height, width) where the 4 axis show:

     1. the cell segmentations filled,
     2. nucleus segmentation filled,
     3. cell segmentation outlines,
     4. nucleus segmentation outlines

     Each pixel in these segmented images will either have value 0, denoting no cell, or the value of the cell ID.

   • A data.csv file

     This file contains information output from Cytokit. Each row represents a cell, and the column rid gives the cell index in the .h5ad file. The column id gives the cell ID for that specific tile. The column z gives which z-plane in which that particular tile has the best focus, and columns 'tile_x' and 'tile_y' give the coordinates of the tile for that cell.

     A folder named images containing the tiff files and the data.csv file need to be compressed into a .tar.gz file and uploaded to Cellar in the spatial panel.

     Here is an example tar.gz file that corresponds to CODEX_Florida_19-003-lymph-node-R2 in Cellar.

Generating the tile will color each cell by its cluster.

Furthermore, tiles can also be colored by protein expression

Visium 10x Spatial Transcriptomics

For 10X spatial data, simply upload the spatial related files generated by spaceranger. For example, for this Adult Mouse Brain (FFPE) dataset, download the “Spatial imaging data” and upload it in the Spatial Data panel.

There is no need to upload any spatial image data for the 10x_breast_cancer_w_image dataset found on our web server.

Note that “Feature / barcode matrix (filtered)” gene expression files can be uploaded to Cellar in the Load Data panel directly.

Also colored by gene expression