communiquer

A simple interface for the tools that predict cells communication from Single-Cell RNA sequencing data

Installation for the development

1. Clone the repository

$ git clone https://github.com/VladimirShitov/communiquer.git

2. Install poetry. Recommended way:

$ curl -sSL https://raw.githubusercontent.com/python-poetry/poetry/master/get-poetry.py | python -

Alternative way (not recommended):

$ pip install --user poetry

3. Install project's python dependencies:

$ poetry install

4. Install tools for inferring cells communications. Follow the instructions from the repositories:

• Cellchat
• CellPhoneDB
• CellCall (not yet supported)

Make sure that cellphonedb is added to the PATH (i. e. you can run it via command line from any directory)

5. Run the code!

Examples

Preparation

Import packages:

from pathlib import Path

from communiquer.launchers import CellChatLauncher, CellPhoneDBLauncher
from communiquer.drawing import draw_chord_diagram

Provide paths to the data. You have to prepare 2 files:

1. Normalized counts matrix. Rows are genes, columns are barcodes (or other cells IDs
2. Meta information about cells. It must have 2 columns: the first one with cell IDs (e.g. barcodes), and the second with cells labels (e.g. cell types)

counts_path = Path("data/pbmc3k_counts.tsv")
meta_path = Path("data/pbmc3k_meta.txt")

cpdb_output_path = Path("data/cpdb_output")

Example of the meta file:

Cell	cell_type
AAACATACAACCAC-1	RPS expressing
AAACATTGAGCTAC-1	B
AAACATTGATCAGC-1	CD4 T
AAACCGTGCTTCCG-1	CD14 Monocytes
AAACCGTGTATGCG-1	NK
...

Running CellPhoneDB

cpdb_launcher = CellPhoneDBLauncher(
    meta_file_path=meta_path,
    counts_file_path=counts_path,
    output_path=cpdb_output_path,  # CellPhoneDB saves its output to the files
    project_name="pbmc3k",
    counts_data="gene_name"  # If you have Ensemble IDs, leave default
)
cpdb_launcher.run()
cpdb_launcher.read_output(convert_to_cellchat_format=True)

Then you can display the output by checking:

cpdb_launcher.pvalues_df
cpdb_launcher.means_df

To see outputs of this code, check example notebook

Count significant interactions and build a chord diagram with them:

cpdb_launcher.count_significant_interactions() # This sets cpdb_launcher.counts_df
cpdb_launcher.visualise_interactions()

Or a dot plot:

Running CellChat

cellchat_launcher = CellChatLauncher(counts_file_path=counts_path, meta_file_path=meta_path)
cellchat_launcher.run()
cellchat_launcher.read_output()

You can then view the output:

cellchat_launcher.weights_df
cellchat_launcher.pvalues_dfs
cellchat_launcher.probabilities_dfs

Cellchat stores every interaction in a different matrix. You can access these interactions by name:

cellchat_launcher.pvalues_dfs["ESAM_ESAM"]

Check the outputs in the example notebook

You can run visualisations with the same interface:

cellchat_launcher.visualise_interactions()

cellchat_launcher.dotplot_counts()