Code for running base calling for PerturbView in tissue, and aligning tissue data to Xenium Spatial Transcriptomics.
Tested on python 3.10.13. Later versions of python may not work
install the requirements in a virtualenv or conda-env with:
pip install -r requirements.txt
The spotiflow general model must also be copied to ./models/general/
The pipeline is currently run through a snakemake script. Please edit the snakefile to point to relevant paths for your experiment:
sub_dir = this is the name of your experiment
sample_dir = location of your ISS imaging. We store this in {sub_dir}/ISS
pheno_dir = location of your phenotyping imaging (antibodies, stains, or Xenium DAPI image). We store this in {sub_dir}/pheno
For calling gRNAs or other barcodes via ISS ensure that a CSV is filed set to the grnas variable. The file should at minimum contain a column called either gene_symbol or ID, refering to the barcode name and another column called spacer which is the full nucleotide string of the region being ISS'd (e.g. AATTGGCC) if fewer nucleotides than the full spacer are sequenced only in situ only the relevant bases will be matched.
Once all variables are set the snakefile can be run by activating the environment than typing
snakemake --cores all
The script is relatively memory efficient, if you find it crashing you may need to cut the number of cores down. Alternatively, if you run code some of the modules on a GPU, you may need to cut concurrency to ensure the GPU RAM isn't overwhelmed.
The script is currently setup to register Ab phenotyping to ISS.
pheno_dir should properly point to your phenotyping images.
Xenium DAPI images can be directly used to register to tissue ISS.
Set pheno_img to the location of your DAPI morphology ome.tiff file. The DAPI Z plane used for registering the data is currently set to Z=0 on line 50 of register_xenium.py. Change this as needed.
To adapt the script to register Xenium to ISS the following rules should also be removed:
stitch_pheno # Xenium is already stitched
segment_phenotype # Xenium is already segmented
bg_subtract_pheno # Not needed
get_nuclei_masks # Xenium is already segmented
Additionally the keyword argument --xenium should be added to the rule call_peaks.
For now, please download the dataset from here. https://ftp.ebi.ac.uk/pub/databases/IDR/idr0162-kudo-perturbview/