NB: This repository aims to provide everything needed to rerun the original Jupyter notebook. However, it comes without the guarantee - implied or expressed - that the information in this repository is sufficient to reproduce the rusults perfectly. I am not connected to the Theis Lab and work on this repository in my free time.
The analyses were performed in Python and heavily use scanpy and anndata functionality. Both Python packages were originally created for single cell transcriptomics analyses, but can to a large part also directly be applied to protein data and thus help kick off this exciting new field! Especially scanpy enables easy data visualizations like PCA plots or UMAPs and provides some ready to use analysis methods like assigning cell cycle stage scores to single cells given some cell cycle markers.
The only requirements for the reproduciton are a computer with a conda installation or a container engine able to work with Dockerfile and this repository.
reproduce-with-conda-and-containers.zip is the release that containins all information necessary for the reproduction of notebook.
- Get a copy of this repository and open a shell in it.
- Set up the conda environment using the conda environment definition in
reproduce-single-cell-proteomics.yaml. This will create an environment namedscprep. - Activate the new conda environment
- Start JupyterLab
- Open
TSP_cell_cycle_analysis.ipynbin JupyterLab - Run the code.
For convenience step 1.-4. as shell commands:
# Step 1. Copy the repo
git clone https://github.com/fabianegli/singlecell_proteomics
# Step 1. Navigate in a terminal into the repo
cd singlecell_proteomics
# Step 2.
conda env create --file reproduce-single-cell-proteomics.yaml
# Step 3.
conda activate scprep
# Step 4. unpack the bundled data
python prepare_data.py
# Step 5.
jupyter labThe reproduction with containers has been tested with podman.
It may also work with other containerization technologies that can build containers from Dockerfile like Docker or Singularity.
e.g. for Docker it should be possible to run the same analysis by simply changeing podman to docker.
Both, interactive and non-interactive methods to reproduce the analysis require a container image.
It can be built with podman build --tag tscp . run in the root of a clone of this repository.
For this to work podman needs to be installed on the system.
The Dockerfile in this repository can be used as follows to run the notebook:
- Unpack the bundled data with
podman run -it -v "${PWD}":"/tscp/" tscp bash -c "cd tscp/; python prepare_data.py" - Run the Podman image with
podman run -it -p 8888:8888 -v "${PWD}":"/tscp/" tscp - Point the browser to link in shown in the terminal starting with
http://127.0.0.1:8888/ - Open the tscp folder in the Jupyter lab navigation, it contians the notebook.
- Open and run the notebook.
For a non-interactive reproduction one can use the following command:
podman run -it -v "${PWD}":"/tscp/" tscp \
bash -c \
"\
cd tscp; \
sed -i 's/conda-env-tissue-py/python/' TSP_cell_cycle_analysis.ipynb; \
python prepare_data.py; \
python -m nbconvert \
--execute \
--to notebook \
TSP_cell_cycle_analysis.ipynb \
--output=TSP_cell_cycle_analysis-reproduced.ipynb \
"
The sed command above is necessary, because the jupyter notebook contains information about the Python environment used by the original authors which is not available.
All data required for successful reproduction and not already available in this repository has be collected and compiled into a data bundle.
From PRIDE project [PXD024043] the files 20210919_DIANN_SingleCellOutput.pg_matrix_cellcyclepred.tsv and 20210919_DIANN_SingleCellOutput.pg_matrix_notnormalized.tsv in the
DIANN1.8_SingleCells_CellCycle.7z archive available from PRIDE project PXD024043 licensed under the Creative Commons Public Domain (CC0).
All other files in data.tar.gz originate from Gene Expression Omnibus (GEO) as downloaded in January 2023.
- directory
GSE129447_RAW - directory
GSE142277_RAW - file
GSM4226257_out_gene_exon_tagged.dge_exonsds_046.txtinGSM4226257_RAW - file
GSM4226257_out_gene_exon_tagged.dge_intronsds_046.txtinGSM4226257_RAW
The contents of data.tar.gz
data
├── 20210919_DIANN_SingleCellOutput.pg_matrix_cellcyclepred.tsv
├── 20210919_DIANN_SingleCellOutput.pg_matrix_notnormalized.tsv
├── GSE129447_RAW
│ ├── GSM3713084_HeLa_1.txt
│ ├── GSM3713085_HeLa_2.txt
│ └── GSM3713086_HeLa_3.txt
├── GSE142277_RAW
│ ├── GSM4224315_out_gene_exon_tagged.dge_exonssf002_WT.txt
│ ├── GSM4224315_out_gene_exon_tagged.dge_intronssf002_WT.txt
│ ├── GSM4224316_out_gene_exon_tagged.dge_exonssf002_KO.txt
│ └── GSM4224316_out_gene_exon_tagged.dge_intronssf002_KO.txt
└── GSM4226257_RAW
├── GSM4226257_out_gene_exon_tagged.dge_exonsds_046.txt
└── GSM4226257_out_gene_exon_tagged.dge_intronsds_046.txt
This repository contains analyses done for Ultra-high sensitivity mass spectrometry quantifies single-cell proteome changes upon perturbation, A Brunner, ..., M Mann, Molecular Systems Biology (2022).
This repository is licensed under the BSD-3 clause license.
The licenses of the bundled data is noted in the "The budled data in data.tar.gz" section above.
The effort to make this repository's code reproducible was carried out on MacOS 11.7.2 (MacBook Pro late 2013) and PureOS (Librem 14). It has yet to be tested on other platforms.
conda-env-for-reproduction-on-MacOS1172.yaml contains the exported conda environment - this file can be used to set up the environment exactly the same, but it will most probably not work if you use another OS than what generated it. Analogously, the podman-conda-env-export.yaml is a record ot the conda environment in a container running on PureOS.
While this repository holds all the information to reproduce the plots, there might still be subtle differences that escaped or that are outside my purview to change. The plots also don't exactly match pixel for pixel - there are some matplotlib/seaborn specific settings which could not be reproduced.