descopex 0.0.7

Decoding Single-Cell Omics of Perturbation Responses Using DeSCOPE

DeSCOPE is a single-cell perturbation prediction framework designed for scRNA-seq, scATAC-seq, and general single-cell–level perturbation modeling. It is built on a conditional Variational Autoencoder (cVAE) architecture, in which perturbed genes are represented by embeddings derived from the ESM2 protein language model and used as conditioning information to model cellular responses to genetic perturbations. Through this design, DeSCOPE delivers strong predictive performance in challenging scenarios, including unseen genes and unseen cell types.

Installation

Environment Setup

Step 1: Set up a python environment

We recommend creating a virtual Python environment with Anaconda:

• Required version: python >= 3.10

conda create -n descope python=3.10
conda activate descope

Step 2: Install pytorch

Install PyTorch based on your system configuration. Refer to PyTorch installation instructions.

For the exact command, for example:

• You may choose any version to install, but make sure the PyTorch version is not too old.
• We recommend torch ≥ 2.6.

# Installation Example: torch v2.7.1
# CUDA 11.8
pip install torch==2.7.1 torchvision==0.22.1 torchaudio==2.7.1 --index-url https://download.pytorch.org/whl/cu118
# CUDA 12.6
pip install torch==2.7.1 torchvision==0.22.1 torchaudio==2.7.1 --index-url https://download.pytorch.org/whl/cu126
# CUDA 12.8
pip install torch==2.7.1 torchvision==0.22.1 torchaudio==2.7.1 --index-url https://download.pytorch.org/whl/cu128

Step 3: Install deepspeed (optional)

Install DeepSpeed based on your system configuration. Refer to DeepSpeed installation instructions.

For the exact command, for example:

pip install deepspeed

Step 4: Install descope and dependencies

To install descope, run:

pip install descopex

Or install from github:

git clone https://github.com/Peg-Wu/DeSCOPE.git
cd DeSCOPE
pip install [-e] .

Check if installation was successful:

import descope
descope.welcome()

Datasets Zoo

scRNA-seq

• The H1 dataset was obtained from the Virtual Cell Challenge 2025. (download)
  • NOTE: ESM2 gene embeddings can also be obtained from here.

scATAC-seq

• Cell-by-cCRE / Cell-by-peak: Human-scATAC-Corpus

Acknowledgements

We sincerely thank the authors of following open-source projects:

Star History