You have the flexibility to execute solely the cell type assignment for your h5ad file, bypassing the need for deconvolution and other processing steps. This customization is achievable within the pipeline framework. To tailor the pipeline to perform only the cell type assignment, you can adjust the input.nf file as follows:
params{
celltype_assignment{
run_celltype_assignment=true
run_keras=true
run_celltype_assignment=false
run_azimuth=true
run_celltypist=false
}
skip_preprocessing=true
file__anndata_merged = '/lustre/scratch126/humgen/projects/sc-eqtl-ibd/analysis/yascp_analysis/2023_08_22-sequencing_pilots/ti_Element/results/merged_h5ad/pre_QC_adata.h5ad'
}
And then you can run the pipeline as:
nextflow run /path/to/cloned/yascp -profile sanger -entry JUST_CELLTYPES -c input.nf
module load HGI/pipelines/yascp/1.5
yascp celltype -c input.nf
Additionally, if you possess custom CellTypist models you wish to utilize, you can modify the default parameters. For an overview of the models currently available by default, please refer to the config file
params{
celltypist {
models = ['/path/to/my/model/Immune_All_High.pkl']
}
}
Additionally, if you have a custom Azimuth model or one obtained from Azimuth Zenodo that you wish to use, you can modify the default parameters. By default, the pipeline uses a PBMC reference, but you have the flexibility to change this and include any references that meet your requirements.
params{
azimuth{
run_process = true
celltype_refsets = [
//# [ name : 'kidney', refset : "/lustre/scratch123/hgi/teams/hgi/mo11/tmp_projects/jaguar_yascp/nieks_pipeline/yascp_run/ref_kidney", annotation_labels : "cluster,subclass" ],
[ name : 'PBMC', refset : "PBMC", annotation_labels : "celltype.l2,celltype.l1,celltype.l3" ],
]
}
}