CellEnBoost: A Boosting-Based Ligand-Receptor Interaction Identification Model for Cell-to-Cell Communication...

Area of Science:

• Computational biology
• Cancer research
• Bioinformatics

Background:

• Cell-to-cell communication (CCC) is crucial for multicellular organisms, influencing cancer genesis, development, and metastasis.
• Ligand-receptor interactions (LRIs) are the primary mediators of CCC, particularly within the complex tumor microenvironment.
• Accurate identification of LRIs is essential for understanding cancer biology and developing targeted therapies.

Purpose of the Study:

• To develop and validate a novel boosting-based computational model, CellEnBoost, for predicting LRIs involved in CCC.
• To apply CellEnBoost for elucidating cell-cell communication patterns in cancer, specifically in human head and neck squamous cell carcinoma (HNSCC).
• To provide a robust tool for inferring and visualizing complex cellular communication networks.

Main Methods:

• Development of CellEnBoost, an ensemble model combining Light gradient boosting machine, AdaBoost, and convolutional neural networks for LRI prediction.
• Integration of data collection, feature extraction, and dimensional reduction techniques for robust LRI identification.
• Application of CCC strength measurement and single-cell RNA sequencing data for LRI filtering and CCC elucidation, with results visualized via heatmap, Circos, and network plots.

Main Results:

• CellEnBoost demonstrated superior performance, achieving the highest Area Under the Curve (AUC) and Area Under the Precision-Recall Curve (AUPR) across four LRI datasets.
• Case study on HNSCC tissues revealed significant communication between fibroblasts and HNSCC cells, consistent with existing methods like iTALK.
• The model successfully identified and visualized key LRIs, providing insights into tumor microenvironment interactions.

Conclusions:

• CellEnBoost is an effective and accurate computational tool for inferring cell-to-cell communication through LRI identification.
• The findings highlight the importance of fibroblast-cancer cell communication in HNSCC, offering potential therapeutic targets.
• This work contributes a valuable approach to cancer research, potentially aiding in diagnosis and treatment strategies.