Introduction

Nextflow pipeline for multiplex imaging spatial data methods.

Spatial PHLEX is a constituent component of TRACERx-PHLEX, a modular pipeline for key data analysis tasks in multiplexed imaging. For a detailed guide to Spatial PHLEX's capabilities, please see the TRACERx PHLEX paper or its documentation.

Requirements

Spatial PHLEX was developed for HPC systems with both CPU and GPU compute facilities. This is because Spatial PHLEX cellular barrier scoring requires GPU-accelerated graph algorithms in order to compute the required shortest paths analysis in a reasonable time. If you do not have GPUs on your system then you can still run Spatial PHLEX in spatial_clustering mode. See Workflow Options.

Inputs

• cell_objects.csv | cell_objects.txt
  • A plaintext, delimited file containing single cell-level coordinate data for a set of images, plus their phenotypic identities.
• metadata.csv (optional)
  • A plaintext, delimited file containing metadata information about the images in cell_objects.csv. To run the pipeline this file must contain, for each image, an image identifier ('imagename'), and the width and height in pixThis option is preferred if it is desirable for mask and plot outputs to match the input image shape exactly.

Workflow options

Spatial-PHLEX can be run in multiple modes:

• 'default', 'clustered_barrier'
  • The default pipeline. This performs density-based spatial clustering of all cell types in the objects dataframe. Subsequently, graph-based barrier scoring is executed for a defined, source (e.g. CD8 T cells), target (e.g. Epithelial cells), and barrier (e.g. Myofibroblasts) cell type.
• 'spatial_clustering'
  • Performs only spatial clustering, for all cell types identified in the cell objects dataframe.
• 'barrier_only'
  • Performs graph-based barrier scoring without performing spatial clustering.

Example data

Example data is provided in the data directory. This data is derived from 5 imaging mass cytometry images from the TRACERx non-small cell lung cancer study, which was first processed with deep-imcyto to generate segmented cells and measure single cell marker intensities, and then the TYPEx phenotyping pipeline to arrive at spatially resolved single cells of discrete types. The data is provided in the following file:

• data/PHLEX_test_cell_objects.csv

Example Usage

First, clone the repository. Then, navigate to the Spatial-PHLEX directory.

git clone https://github.com/FrancisCrickInstitute/Spatial-PHLEX
cd Spatial-PHLEX

Then load Nextflow and Singularity on your system with e.g. (on SLURM):

ml Nextflow/22.04.0
ml Singularity/3.6.4

Set the Singularity cache directory environment variable, to store the container image:

export NXF_SINGULARITY_CACHEDIR='./singularity'

The Spatial PHLEX pipeline can then be run with the following command:

nextflow run ./main.nf \
    --workflow_name 'clustered_barrier' \
    --objects "./data/PHLEX_test_cell_objects.csv"\
    --objects_delimiter "\t" \
    --image_id_col "imagename"\
    --phenotyping_column 'majorType'\
    --phenotype_to_cluster 'Epithelial cells'\
    --x_coord_col "centerX"\
    --y_coord_col "centerY"\
    --barrier_phenotyping_column "majorType" \
    --barrier_source_cell_type "CD8 T cells"\
    --barrier_target_cell_type "Epithelial cells"\
    --barrier_cell_type "aSMA+ cells"\
    --n_neighbours 10\
    --outdir "../results" \
    --release 'PHLEX_test' \
    --plot_palette "./assets/PHLEX_test_palette.json" \
    -w "./scratch"\
    -profile {your_nf-core_profile} \
    -resume

Replace {your_nf-core_profile} with the profile for your organisation, e.g. crick. For more information on nf-core profiles, see the list of available nf-core configs plus other advice on pipeline configuration here.

Outputs

Credits

Spatial PHLEX is primarily developed by Alastair Magness at The Francis Crick Institute. Other core contributors include Emma Colliver, Mihaela Angelova, and Katey Enfield.