We perform an integrative analysis of multiplexed proteomics single cell spatial data from breast cancer tissues using deep convolutional variational autoencoders.
The repository does not contain any biological data and the data is not publically available.
The code should be run on the DKFZ cluster since the data is stored in our private group folder, under /icgc/dkfzlsdf/analysis/B260/projects/spatial_zurich/data
To perform an exploratory data analysis, first login into the DFKZ cluster with the -X option.
After cloning the repository, install the dependencies with
conda env create -f spatial_ops-dev-requirements.yml
If running the code (see later) some dependencies are still missing, this would probably mean that the requirements of the project have been changed since the time you created the environment, so you need to run
conda env update -f spatial_ops-dev-requirements.yml
Now activate the conda environment
conda activate spatial-dev
and simply run snakemake
snakemake
If you are not in the cluster you first need to update the code in folders.py by inserting the path of the root folder of the data in your machine (note that the data is not publically available). In the root folder the data must be organized into this directory tree:
<data_root_folder>/
├── csv/
│ ├── Basel_PatientMetadata.csv
│ ├── Basel_Zuri_SingleCell.csv
│ ├── Basel_Zuri_StainingPanel.csv
│ ├── Basel_Zuri_WholeImage.csv
│ └── Zuri_PatientMetadata.csv
├── Basel_Zuri_masks/
│ └── *.tiff (746 files)
└── ome/
└── *.tiff (746 files)
The data, from the B. Bodenmiller lab, is a collection of images acquired with Imaging Mass Citometry of breast cancer cells of different patients and under different conditions [1].
Each .tiff file in the ome folder is uniquely paired with a .tiff mask. Each mask tells which are the cells.
| [1] | Schulz D, Zanotelli VRT, Bodenmiller B. et al. Simultaneous Multiplexed Imaging of mRNA and Proteins with Subcellular Resolution in Breast Cancer Tissue Samples by Mass Cytometry. Cell Syst. 2018 Jan 24 |