CCC-Guided Latent Representation Learning Enhances Cell-Cell Communication Analysis in Single-Cell RNA-seq Data

Requirements

Python: 3.9.13
PyTorch: 2.6.0 (https://pytorch.org)
Scanpy: 1.10.3 (https://scanpy.readthedocs.io/en/stable)
Numpy: 1.26.4 (https://numpy.org)
Pandas: 2.2.3 (https://pandas.pydata.org)
h5py: 3.13.0 (https://pypi.org/project/h5py)
Seaborn: 0.13.2 (https://seaborn.pydata.org)

Usage

Download the data from https://drive.google.com/drive/folders/1UlasBJkr6PuV0GKmi6sbTuog67FpEd-r?usp=sharing.

Run the process.py under the datasets folder to prepare data.

cd dataset

python process.py

cd experiments

python step_1_calculate_CCC.py --dataset_name pbmc4k --HVG_NUM 10000 --ccc_db consensus

python step_2_train_VAE.py --dataset_name pbmc4k --HVG_NUM 10000 --ccc_db consensus

python step_3_evaluate.py --dataset_name pbmc4k --HVG_NUM 10000 --ccc_db consensus

The dataset_name options are ["pbmc4k", "pbmc12k", "Pancreas", "Opium"], and the available ccc database can be found in the supplementary folder. Other hyperparameters are available in the script file.