Deconvoluting spatial data

With the deconvolution_spatial workflow, one or multiple spatial slides can be deconvoluted in one run. For that, a SpatialData object for each slide is expected. The spatial slides are deconvoluted using the same reference. For the reference, one MuData with the gene expression data saved in mdata.mod["rna"] is expected as input.

The workflow provides the possibility to run deconvolution using Cell2Location and Tangram.

Steps

Cell2Location

For the reference and each spatial slide the following steps are run. Note, that if multiple slides are deconvoluted in one run, the same parameter setting is used for each slide.

  • Gene selection. There are two possibilities for the gene selection:

    1. Genes of a user-provided feature set (csv-file) are used for deconvolution. All genes of that gene list need to be present in both, spatial slides and scRNA-Seq reference.

    2. Feature selection performed according to Cell2Location, i.e. via the function: cell2location.utils.filtering.filter_genes

    Note: if no csv-file is provided by the user, the workflow will run the feature selection via the function: cell2location.utils.filtering.filter_genes. Thus, gene selection is not optional.

  • Regression/reference model is fitted and a plot of the training history as well as QC plots are saved in the ./figures/Cell2Location directory. Additionally, a csv-file Cell2Loc_inf_anver.csv with the estimated expression of every gene in every cell type is saved in ./cell2location.output.

  • (Optional) Reference model is saved in ./cell2location.output

  • Spatial mapping model is fitted. Training history and QC plots are saved in the ./figures/Cell2Location directory. Plot of the spatial embedding coloured by q05_cell_abundance_w_sf is also saved in ./figures/Cell2Location.

  • (Optional) A gene by spot matrix for each cell type is saved to a layer in the table of the SpatialData object

  • (Optional) Spatial mapping model is saved in ./cell2location.output

  • The SpatialData object of the spatial slide and the MuData object of the reference are saved in ./cell2location.output. The SpatialData object of the spatial slide contains the estimated cell type abundances.

Tangram

For the reference and each spatial slide the following steps are run. Note, that if multiple slides are deconvoluted in one run, the same parameter setting is used for each slide.

  • Gene selection with two possibilities:

    1. Genes of a user-provided feature set (csv-file) are used for deconvolution. All genes of that gene list need to be present in both, spatial slides and scRNA-Seq reference.

    2. sc.tl.rank_genes_groups is used to select genes. The top n genes of each group make up the reduced gene set.

    Note: if no csv-file is provided by the user, the workflow will run the feature selection via sc.tl.rank_genes_groups. Thus, gene selection is not optional.

  • Data is preprocessed with tangram.pp_adatas

  • Tangram model is fitted with tangram.mapping_utils.map_cells_to_space and annotations are transfered from single-cell data onto space with tangram.project_cell_annotations

  • Plot of the spatial embedding coloured by tangram_ct_pred is saved in ./figures/Tangram

  • The SpatialData object of the spatial slide and the MuData object of the reference are saved in ./tangram.output. The SpatialData object of the spatial slide contains the deconvolution predictions.

Steps to run

  1. Activate conda environment conda activate pipeline_env

  2. Generate yaml and log file panpipes deconvolution_spatial config

  3. Specify the parameter setting in the pipeline.yml file

  4. Run complete deconvolution workflow with panpipes deconvolution_spatial make full --local

The Deconvoluting spatial data tutorial guides you through deconvolution workflow of Panpipes step by step.