CellGiQ, a novel framework for deciphering ligand-receptor-mediated cell-cell communication by incorporating machine learning and a quartile scoring strategy from single-cell RNA sequencing data. CellGiQ accurately inferred intercellular communication within human HNSCC tissues. CellGiQ is anticipated to dissect cellular crosstalk and signal pathways at single cell resolution.

• python == 3.8.13

• tensorflow == 2.10.0

• keras == 2.10.0

• GBNN == 0.0.2

• interpret == 0.2.7

• scikit-learn == 0.24.0

• lightgbm == 3.3.5

• wheel == 0.37.1

• pands == 1.5.0

• numpy == 1.24.2

1.Data is available at uniprot, GEO.

2.Feature extraction website at BioTriangle

1. We obtain ligand and receptor feature at BioTriangle

2. Run the model to obtain the LRI, or the user-specified LRI database

   python code/CellGiQ.py
   

3. Using quartile method (including Expression thresholding, Expression product and Specific expression), the cell-cell communication matrix was finally obtained.

   python code/case study
   

If you want to test other tumors, just replace GSE103322.csv in the code case_study.py (Note: use the specified database to replace the datasetLRI_dataset.csv)

CellChat iTALK LIANA CellPhoneDB NATMI