imcpipeline 0.0.5

A pipeline and utils for IMC data analysis.

Imaging mass cytometry pipeline

This is a pipeline for the processing of imaging mass cytometry (IMC) data.

It is largely based on Vito Zanotelli's pipeline. It performs image preprocessing and filtering, uses ilastik for semi-supervised pixel classification, CellProfiler for image segmentation and quantification of single cells.

The pipeline can be used in standalone mode or with imcrunner in order to process multiple samples in a distributed way and in parallel such as a local computer, on the cloud, or a high performance computing cluster (HPC). This is due to the use of the light-weight computing configuration manager divvy.

Requirements and installation

Requires:

• Python >= 3.7
• One of: docker, singularity, conda or cellprofiler in a local installation.

Install with:

pip install imcpipeline

Make sure to have an updated PIP version. Development and testing is only done for Linux. If anyone is interested in maintaining this repository in MacOS/Windows fell free to submit a PR.

Quick start

Demo

You can run a demo dataset using the --demo flag:

imcpipeline --demo

The pipeline will try to use a local cellprofiler installation, docker or singularity in that order if any is available. Output files are in a imcpipeline_demo_data directory.

Running on your data

To run the pipeline on real data, one simply needs to specify input and output directories. A trained ilastik model can be provided and if not, the user will be prompted to train it.

imcpipeline \
    --container docker \
    --ilastik-model model.ilp \
    -i input_dir -o output_dir

If docker or singularity is not available, one could for example use a conda environment or a virtualenv environment activated only for the cellprofiler command like this:

imcpipeline \
    --cellprofiler-exec \
        "source ~/.miniconda2/bin/activate && conda activate cellprofiler && cellprofiler"
    --ilastik-model model.ilp \
    -i input_dir -o output_dir

To run one step only for a single sample, use the -s/--step argument:

imcpipeline \
    --step segmentation \
    -i input_dir -o output_dir

Or provide more than one consecutive step in the same way:

imcpipeline \
    --step predict,segmentation \
    -i input_dir -o output_dir

To run the pipeline for various samples in a specific computing configuration (more details in the documentation):

imcrunner \
    --divvy-configuration slurm \
    metadata.csv \
        --container docker \
        --ilastik-model model.ilp \
        -i input_dir -o output_dir

Documentation

For additional details on the pipeline, see the documentation.

Related software

• Vito Zanotelli's pipeline;
• A similar pipeline implemented in Nextflow.