SEnSCA: Identifying possible ligand-receptor interactions and its application in cell-cell communication inference

Area of Science:

• Computational Biology
• Bioinformatics
• Systems Biology

Background:

• Multicellular organisms rely on complex cell-cell communication (CCC) for coordinated activity.
• Current methods for inferring CCC are limited to known ligand-receptor interactions (LRIs).

Purpose of the Study:

• To develop a comprehensive computational tool, SEnSCA, for analyzing CCC by integrating single-cell RNA sequencing and proteome data.
• To overcome the limitations of existing methods by enabling the discovery of potential LRIs.

Main Methods:

• SEnSCA employs a Stacking ensemble model (SVM, 1D-CNN, multi-head attention) for potential LRI acquisition after feature extraction and dimension reduction.
• Negative LRI samples are generated using K-means clustering.
• CCC strength is evaluated by LRI filtering and a three-point estimation approach combined with single-cell RNA sequencing data.

Main Results:

• SEnSCA demonstrated superior performance in predicting potential LRIs, achieving higher precision, recall, accuracy, F1 score, AUC, and AUPR compared to existing methods.
• The tool provides three visualization options (heatmap, bubble diagram, network diagram) for clear representation of CCC networks.
• Application to human melanoma tissue confirmed SEnSCA's reliability in CCC detection.

Conclusions:

• SEnSCA offers a robust and versatile tool for inferring cell-cell communication networks.
• The method enhances the discovery of potential ligand-receptor interactions, advancing the understanding of cellular coordination.
• SEnSCA is freely available, promoting wider accessibility for biological research.