Welcome to CellDISECT’s documentation!

Note

Beta Version Available: A beta version (0.2.0b1) with compatibility for Google Colab and newer versions of torch and scvi-tools is available on the beta-colab branch. Install it with pip install celldisect==0.2.0b1.

CellDISECT (Cell DISentangled Experts for Covariate counTerfactuals) is a causal generative model designed to disentangle known covariate variations from unknown ones at test time while simultaneously learning to make counterfactual predictions.

Contents

• Installation
  • 1. Create and activate a conda environment:
  • 2. Install PyTorch (tested with pytorch 2.1.2 and cuda 12):
  • 3. Install CellDISECT:
  • Optional Dependencies
• Example Scripts
  • Training Example
  • Inference Example
  • Using the Examples
  • Perturbation Example
• Tutorials
  • Getting Started
  • Tutorial Categories
  • Detailed Tutorial Contents
  • Tutorial Details
• API
  • Model
  • Data
  • Utilities
  • Model
• Changelog
  • Beta Versions
  • 0.2.0b1 (2024-03-16)
  • Stable Versions
• Contributing
  • Development Process
  • Pull Request Process
  • Any contributions you make will be under the BSD 3-Clause License
  • Report bugs using GitHub’s issue tracker
  • Write bug reports with detail, background, and sample code
  • Development Setup
  • License
• References
• Other References

Installation

Prerequisites

We recommend using Anaconda/Miniconda to create a conda environment for using CellDISECT.

1. Create and activate a conda environment:

conda create -n CellDISECT python=3.9
conda activate CellDISECT

2. Install PyTorch (tested with pytorch 2.1.2 and cuda 12):

conda install pytorch==2.1.2 torchvision==0.16.2 torchaudio==2.1.2 pytorch-cuda=12.1 -c pytorch -c nvidia

3. Install CellDISECT:

You can install the stable version using pip:

pip install celldisect

Or install the latest development version from GitHub:

pip install git+https://github.com/Lotfollahi-lab/CellDISECT

Optional Dependencies

For RAPIDS/rapids-singlecell support:

pip install \
    --extra-index-url=https://pypi.nvidia.com \
    cudf-cu12==24.4.* dask-cudf-cu12==24.4.* cuml-cu12==24.4.* \
    cugraph-cu12==24.4.* cuspatial-cu12==24.4.* cuproj-cu12==24.4.* \
    cuxfilter-cu12==24.4.* cucim-cu12==24.4.* pylibraft-cu12==24.4.* \
    raft-dask-cu12==24.4.* cuvs-cu12==24.4.*

pip install rapids-singlecell

For CUDA-enabled JAX:

pip install -U "jax[cuda12_pip]" -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html

Quick Start

Here’s a simple example to get you started:

from celldisect import CellDISECT
import scanpy as sc

# Load your data
adata = sc.read_h5ad('your_data.h5ad')
adata.X = adata.layers['counts'].copy()
cats = ['cov1', 'cov2']
cell_type_included = False
# Initialize and train the model
CellDISECT.setup_anndata(
 adata,
 layer='counts',
 categorical_covariate_keys=cats,
 continuous_covariate_keys=[],
 add_cluster_covariate=not cell_type_included, # add_cluster_covariate if cell type is not included
)
model = CellDISECT(adata)
model.train()

# Make predictions
predictions = model.predict_counterfactuals(
    adata,
    cov_names=['cov1'],
    cov_values=['val1'],
    cov_values_cf=['val2'],
    cats=cats,
)