Cell-cell communication inference and analysis: biological mechanisms, computational approaches, and future opportunities

TL;DR

This review tackles the challenge of inferring and analyzing cell-cell communication (CCC) from single-cell and spatial omics data. It categorizes and synthesizes more than 140 computational methods into five core families—statistical, network, deep learning, optimal transport, and factorization—highlighting how each addresses spatial constraints, single-cell resolution, intracellular signaling, temporal dynamics, and cross-condition comparisons. Key contributions include a comprehensive mapping of methodological diversity, visualization and systems-analysis tools, and critical discussion of benchmarking, ground-truth validation, and opportunities for de novo CCC construction and mechanistic modeling. The authors emphasize the importance of integrating non-protein signaling, multi-omics data, and clinical information to enable robust CCC inference and to drive engineering and therapeutic applications.

Abstract

In multicellular organisms, cells coordinate their activities through cell-cell communication (CCC), which are crucial for development, tissue homeostasis, and disease progression. Recent advances in single-cell and spatial omics technologies provide unprecedented opportunities to systematically infer and analyze CCC from these omics data, either by integrating prior knowledge of ligand-receptor interactions (LRIs) or through de novo approaches. A variety of computational methods have been developed, focusing on methodological innovations, accurate modeling of complex signaling mechanisms, and investigation of broader biological questions. These advances have greatly enhanced our ability to analyze CCC and generate biological hypotheses. Here, we introduce the biological mechanisms and modeling strategies of CCC, and provide a focused overview of more than 140 computational methods for inferring CCC from single-cell and spatial transcriptomic data, emphasizing the diversity in methodological frameworks and biological questions. Finally, we discuss the current challenges and future opportunities in this rapidly evolving field.

Figures (7)

• Figure 1: Diverse types of cell-cell communication. Paracrine signaling involves signaling molecules that act on neighboring cells within a localized area by diffusing through the extracellular space. Autocrine signaling refers to signals that act on the same cell that produces them or on a population of identical cell types. Contact-dependent signaling relies on direct physical contact between neighboring cells, such as direct membrane contact or gap junctions. Synaptic signaling enables long-distance communication through the specialized synaptic structures of neurons. Endocrine signaling involves signaling molecules produced by signaling cells that reach distant target cells via extracellular fluids, such as blood. Additionally, extracellular vesicles (EVs) also mediate intercellular signaling by delivering a range of bioactive molecules (e.g., proteins, lipids, nucleic acids) through paracrine, autocrine, and endocrine mechanisms. This figure is inspired by a previous workalberts2022molecular. Created in https://BioRender.com
• Figure 2: Cellular signal transduction mediated by different receptor types and major modes of intracellular signaling crosstalk. (a) Cell-surface receptor signaling: Ligand binding activates a cascade of cytoplasmic relay proteins and transcription factors, resulting in the nuclear translocation of transcription factors to regulate target gene expression. (b) Cytoplasmic receptor signaling: The ligand diffuses across the plasma membrane and binds its receptor in the cytoplasm. The resulting ligand-receptor complex then translocates to the nucleus, where it binds specific DNA regulatory sequences to direct transcriptional regulation. (c) Nuclear receptor signaling: The ligand diffuses across the plasma membrane and into the nucleus, where it binds a nuclear receptor. The ligand-receptor complex then directly modulates gene transcription. (d) Single-pathway activation: A lone intracellular signaling pathway is initiated by ligand-receptor binding. (e) Multiple-pathway activation: A single ligand-receptor pair triggers several distinct signaling pathways. (f) Signal convergence: Multiple, distinct ligand-receptor pairs activate a common downstream signaling pathway. (g) Pathway crosstalk: Interdependent signaling pathways, triggered by different ligand-receptor pairs, mutually influence one another. This figure is inspired by a previous workAvior2013 and Pearson Education, Inc. Created in https://BioRender.com
• Figure 3: General strategies and computational methodologies for inferring CCC from gene expression data. (a) A typical workflow for CCC inference. Gene expression profiles from single-cell or spatial transcriptomics serve as proxies for protein abundance. These data are integrated with prior knowledge, including genes encoding interacting ligands and receptors, co-factors, transcription factors and target genes, to infer CCC. A scoring function quantifies the likelihood or strength of potential CCC events, followed by statistical testing to assess significance. The final output predicts CCC at single-cell or cell-group resolution. (b) Common computational methodologies. The five primary strategies for CCC inference include statistical methods, network methods, deep learning, optimal transport, and factorization methods. Created in https://BioRender.com
• Figure 4: Timeline and growth of CCC inference methods. (a) The timeline of 143 CCC inference methods organized by publication or preprint year. Tools based on different methodological principles are distinguished using colored boxes, including statistical methods, network methods, deep learning methods, optimal transport methods, and factorization methods. Symbols denote the specific biological questions addressed. Methods without any symbols are general tools for inferring CCC between cell groups from scRNA-seq data. (b) Cumulative growth of CCC methods. The plot shows the cumulative number of CCC tools developed for scRNA-seq and spatial transcriptomics data over time. The number of methods published up to October 2025 is indicated.
• Figure 5: Methodological phylogeny of CCC tools. The evolutionary tree classifies 143 computational methods by their core computational strategies (main branches, colored) and the specific biological questions they address (sub-branches). This visualization, inspired by a previous reviewRN447, illustrates the evolutionary relationships and functional diversity of CCC tools.