SEnSCA, an innovative framework for unraveling the intricate network of cell-cell communication mediated by ligand-receptor interactions, integrates state-of-the-art machine learning algorithms and a three-point estimation methodology based on single-cell RNA sequencing data. SEnSCA accurately inferred intercellular communication within human melanoma tissues. It is anticipated to dissect cellular crosstalk and signal pathways at single cell resolution.

Install Python 3.9.16 for running this code. And these packages should be satisfied:

• numpy=1.24.3
• pandas=1.5.3
• scikit-learn=1.2.2
• torch=2.0.1
• matplotlib=3.7.1

1. Ligand-receptor data is available at Uniprot, GEO.
2. Feature extraction website at iFeature.

• Step 1: Begin by running the code to identify reliable negative samples, with the option to manually adjust the pre value within the range of 0 to 1.

python Code/Kmeans.py

• Step 2: Execute the code using the default 5-fold cross-validation, and get the ligand-receptor pairs.

python Code/SEnSCA.py

• Step 3: Implement the three-point estimation method, incorporating [cell expression, expression product, specific expression]. (Note: the user-specified database only needs to replace the LRI.csv file and the corresponding format in the file.)
• Step 4: Finally, output the strength of cell-cell communication.

CellChat iTALK CellPhoneDB NATMI CellComNet CellDialog Cellinker SingleCellSignalR Connectome Cytotalk