Skip to content

Latest commit

 

History

History
142 lines (90 loc) · 7.42 KB

README.md

File metadata and controls

142 lines (90 loc) · 7.42 KB
Raw

Constructing cell type PPI from Tabula Sapiens

Step 0: Set up environment & data variables

This codebase leverages Python, NetworkX, Scanpy, etc. To create an environment with all of the required packages, please ensure that conda is installed and then execute the commands:

conda env create -f scRNA_env.yml
conda activate scRNA_env

The data variables for the single cell transcriptomic atlas, global PPI network, auxiliary data, and output directory are in data_config.py. Be sure to update them accordingly.

Step 1: Rank and extract genes

Required environment for each script used:

  • 0.constructPPI.py depends on conda activate scRNA_env

Detailed instructions

First, run 0.constructPPI.py to construct cell type specific PPI networks. It is possible to split the task into two: (1) Ranking the genes in each cell type, and (2) extracting the cell type specific PPI network from the global PPI network based on the rankings.

To rank genes:

python 0.constructPPI.py -rank True -rank_pval_filename PATH/TO/OUTPUT

The outputs of this command are:

  • AnnData object with ranked genes information (.h5ad file type)
  • Table (.csv, tab-delimited) of the ranked genes and their corresponding p-values (rows) for each cell type (columns)

To extract genes:

python 0.constructPPI.py -rank False -rank_pval_filename PATH/TO/INPUT -celltype_ppi_filename PATH/TO/OUTPUT -max_pval 1 -max_num_genes 4000

The outputs of this command are:

  • Table (.csv, tab-delimited) of the cell types (two columns: one for index of cell type, one for name of cell type) and their genes (one column, comma-delimited string of proteins)

How to use each flag:

  • Use the -rank flag to indicate whether you want to rank the genes (True) or extract the PPI network (False).
  • Indicate the column of annotations (in the input AnnData) using the -annotation flag.
  • To subsample equal number of cells per cell type, set the -subsample flag to True and indicate the number of iterations using the -iterations.
  • When -rank False, the relevant flags to include are -max_pval (float, default is 1) and -max_num_genes (integer, default is 3000).
  • Provide the path to the cell type PPI network using the -celltype_ppi_filename flag.

Important parameters that are NOT flags:

  • The global TABULA_SAPIENS_DIR variable holds the directory to AnnData object (.h5ad file type). The AnnData must already have cell type anotations. The default column is cell_ontology_class. Be sure to update TABULA_SAPIENS_DIR with the path to your desired input data.
  • Cell types with fewer than num_cells_cutoff cells are excluded. By default, num_cells_cutoff = 100.

Step 2: Evaluate cell type specific PPI networks

Required environment for each script used:

  • 1.evaluatePPI.py depends on conda activate scRNA_env

Detailed instructions

Next, run 1.evaluatePPI.py to evaluate the cell type specific PPI networks.

python 1.evaluatePPI.py -celltype_ppi PATH/TO/INPUT

The outputs of this command are:

  • PDFs of figures produced during evaluation

How to use each flag:

  • Provide the path to the cell type PPI networks (output from the previous step) using the -celltype_ppi flag
  • Indicate the appropriate -max_pval (float, default is 1) and -max_num_genes (integer, default is 4000). These must be the same values as those used to run 0.constructPPI.py.

Step 3: Run CellPhoneDB for cell-cell interactions

Required environment for each script used:

  • 2.prepCellPhoneDB.py depends on conda activate scRNA_env
  • 3.run_cellphonedb.sh depends on conda activate cpdb
    1. conda create -n cpdb python=3.7
    2. conda activate cpdb
    3. pip install cellphonedb

Detailed instructions

Run 2.prepCellPhoneDB.py to prepare inputs for CellPhoneDB.

python 2.prepCellPhoneDB.py -data PATH/TO/H5AD/DATA -output PATH/TO/OUTPUT

The script outputs are:

  • Meta data (tab-delimited table with two columns: first column is the cell ID, second column is the cell type annotation).
  • (Optional) Count matrix (tab-delimited, columns are cells and rows are genes).

How to use each flag:

  • Use the -data flag to provide the directory of the AnnData object (.h5ad file type). The AnnData must already have cell type anotations. The default column is cell_ontology_class.
  • For the count matrix, CellPhoneDB can either take .h5ad (recommended) or a .txt with the dense matrix. By default, the script does NOT create a .txt file. But if that is desired, set -get_counts True.
  • Provide an output directory to save the files produced by the script using the -output flag.

To run CellPhoneDB, see 3.run_cellphonedb.sh:

cellphonedb method statistical_analysis meta_CellPhoneDB.txt ranked_TabulaSapiens.h5ad --counts-data hgnc_symbol --output-path CellPhoneDB_results

The outputs of this command are:

  • See CellPhoneDB documentation for most up to date information
  • The means.txt file contains mean values for each ligand-receptor interaction. Each row is cell pair interaction.
  • The significant_means.txt file (similar structure as means.txt) contains significant means for all interacting partners. If p-value < 0.05, the value will be the mean. Alternatively, the value is set to 0. This file also contains the rank, which is the total number of significant p-values for each interaction divided by the number of cell type to cell type comparisons.
  • The pvalues.txt file (similar structure as means.txt) contains p-values for the all the interacting partners. THe p-value refers to the enrichment of the interacting ligand-receptor pair in each of the interacting pairs of cell types.
  • The deconvoluted.txt file gives additional information for each of the interacting partners. This is important as some of the interacting partners are heteromers. In other words, multiple molecules have to be expressed in the same cluster in order for the interacting partner to be functional.

Step 4: Construct cell-cell interaction network

Required environment for each script used:

  • 4.constructCCI.py depends on conda activate scRNA_env

Detailed instructions

Run 4.constructCCI.py to summarize CellPhoneDB outputs and construct a cell-cell interaction network.

python 4.constructCCI.py -cpdb_output PATH/TO/CELLPHONEDB/pvalues.txt -cci_edgelist PATH/TO/OUTPUT

The script outputs are:

  • Edgelist of cell-cell interaction network (tab-delimited)

How to use each flag:

  • Provide the directory to the CellPhoneDB outputs using the -cpdb_output flag. This script will automatically extract all pvalues.txt files from the directory
  • Provide the filename for the cell-cell interaction network edgelist using -cci_edgelist flag.

Step 5: Construct meta graph

Required environment for each script used:

  • 5.constructMG.py depends on conda activate scRNA_env

Detailed instructions

Run 5.constructMG.py to extract tissue-tissue associations and combine all components of the meta graph.

python 5.constructMG.py -celltype_ppi PATH/TO/CELLTYPE/PPI -cci_edgelist PATH/TO/CCI/EDGELIST -mg_edgelist PATH/TO/OUTPUT

The script outputs are:

  • Edgelist of meta graph (tab-delimited)

How to use each flag:

  • Provide the path to the cell type PPI networks using the -celltype_ppi flag.
  • Provide the filename for the cell-cell interaction network edgelist using -cci_edgelist flag.
  • Provide the filename for the meta graph edgelist using -mg_edgelist flag.