Data preprocessing¶
This section contains description of preprocessing of Stereo-seq data.
Stereo-seq Analysis Workflow¶
Stereo-seq sequencing data were preprocessed using SAW to generation spatial gene expression matrices in GEM format.
Image files, which are usually in TIFF format, generated during the stereo-seq library construction process should be ready for further process together with the GEM files.
Note
Data of each section were packed with a sinlge Tissue Section ID in STOmicsDB database, including staining image (.tif), GEM file (.gem) as well as relevant annotation information (.txt), and can be accessed with [STTS0000461 - 515].
GEM file and TIFF file for each chip were manually cropped according to the ROI region to extract the expression and image data of each individual in the tissue section.
Image registration¶
Each pair of GEM file and TIFF file was then went through image registration following MIRROR pipeline.
input:
gem file
image file
output:
affine transformation matrix
Cell boundary detection¶
Each ssDNA image was went through object detection and cell segmentation following Detect cell boundaries based on ssDNA images.
input:
image file
output:
cell mask matrix
DNB aggregation¶
With cell mask and affine transformation matrix, the original gem file would be aggregated into putative cells following cell aggregation.
input:
affine transformation matrix
cell mask matrix
gem file
output:
cell-gene expression matrix in h5d format
Cell clustering¶
The generated Cell-by-gene matrix of each individual was processed following the Seurat integration workflow, from which the clustering results were manually annotated to obtain lineage information of each cluster.
input:
cell-gene matrix
output:
cell-lineage annotation
3D alignment¶
Serial annotation images were aligned by the similarities of morphology and annotation color code pair by pair folling SEAM Pipeline.
input:
cell-lineage annotation
images
output:
aligned 3D coordinates
3D mesh building¶
With the aligned 3D coordinates and cell annotation, tissue level triangular meshes were constructed folling Mesh generation pipeline.
input:
3D coordinates
output:
triangular meshes