Hi-C3: a statistical inference-based model for reconstructing higher-order cell-cell communication networks.

Area of Science:

• Computational Biology
• Systems Biology
• Genomics

Background:

• Multicellular organisms require coordinated cell-cell communication (CCC) for development and function.
• Current methods using single-cell RNA sequencing (scRNA-seq) primarily infer pairwise ligand-receptor interactions (LRIs).
• Biological processes often involve coordinated actions of multiple cell types, necessitating models beyond pairwise interactions.

Purpose of the Study:

• To develop a statistical framework, Hi-C3, for inferring both pairwise and higher-order CCC patterns from scRNA-seq data.
• To model receptor expression influenced by collective signaling from multiple ligand-producing cell types.
• To identify key cellular communication hubs within complex multicellular networks.

Main Methods:

• Developed Hi-C3, a statistical inference framework using principles of network diffusion and epidemic dynamics.
• Modeled receptor expression as a Poisson-distributed variable regulated by collective signaling.
• Employed a likelihood-based expectation-maximization (EM) algorithm and a modified PageRank algorithm for network analysis.

Main Results:

• Hi-C3 successfully inferred pairwise CCC comparable to state-of-the-art methods.
• The framework uniquely uncovered complex higher-order CCC structures.
• Identified key cellular communication hubs supported by independent biological and spatial evidence in Arabidopsis thaliana and colorectal cancer datasets.

Conclusions:

• Hi-C3 provides a powerful statistical model and computational framework for uncovering higher-order CCC.
• The method reveals complex multicellular signaling structures often missed by pairwise inference.
• Offers novel insights into cellular organization and communication networks in development and disease.