PreciCE

A unified workflow for data-driven precision cell fate engineering via highly multiplexed gene control

Installation

Using pip:

Ensure you have Python and pip installed on your system. Then run the following command:

pip install -r model/requirements.txt

Using conda:

Ensure you have Anaconda or Miniconda installed on your system. Create a new conda environment (optional but recommended):

conda create --name cell_reprogram_env
conda activate cell_reprogram_env

Install the dependencies:

conda install --file model/requirements.txt

Sample workflow

Worfklow for processing new dataset and running the model

Step 1. Initialize PreciCE data processing workflow and preprocess dataset

path = '../data/Friedman.h5ad'
raw_adata = sc.read_h5ad('../data/Friedman.h5ad')
workflow = precice(adata=raw_adata, 
                   path = path, 
                   cell_filter=True)

Step 2. Run batch correction (optional)

workflow.set_up_scvi(batch_key='day')

## Visualize effects

workflow.scvi_plot_setup()
sc.pl.umap(workflow.adata, color='day')

Step 3. Compute differential expression

### Identify source and target cells
source_name = 'stem'
target_name = 'meso'

workflow.get_DE(source_name=source_name, target_name=target_name)

Step 4. Network inference: Either load pre-existing network

workflow.get_network(cell_type='embryonic stem cell')

Or infer network and edge weights [Can take several hours]

## Set up PySCENIC for given dataset
workflow.set_up_pyscenic(species)

## Run PySCENIC (Takes several hours)
workflow.run_pyscenic()

## Post processing of learnt transcriptional network
workflow.learn_weights()

Step 5. Run PrecICE to identify optimal transcription factor perturbations

transition = source_name +'_to_' + target_name
python_path = '/user/bin/python'
workflow.run_precice(species='human', python_path=python_path,
                     network_path=workflow.network_path,
                     DE_path=workflow.DE_filenames[transition])

Step 6. Plot results: ranked list of perturbations and associated precision score with each perturbation

from plot import make_precision_plot
result_file = %% insert path to results generated by precice

make_precision_plot(result_file, source=source_name, target=target_name)