Benchmarking tools for deciphering cellular crosstalk in spatially-resolved transcriptomics

Cell-cell communication via ligand-receptor signaling is a fundamental mechanism shaping multicellular organization and functional heterogeneity within tissue microenvironments. Recent advances in spatial transcriptomics (ST) have enabled unprecedented opportunities to systematically infer such interactions under the native spatial context. While prior studies have summarized or compared existing cell-cell interaction (CCI) inference methods, comprehensive benchmarking of tools specifically developed for ST applications remains limited.

Here, we present a comprehensive evaluation of nine computational CCI inference methods on a series of realistic simulation settings and nine real ST datasets from three independent studies, spanning Visium, Stereo-seq, and Xenium platforms. Method performance was assessed based on ligand-receptor prediction accuracy, spatial coherence of interactions, biological relevance via pathway enrichment, and computational efficiency.

Our results demonstrate substantial variability in tool performance across spatial resolutions, tissue contexts, and platforms, offering practical guidance for tool selection. This study also highlights key challenges in applying existing tools to real ST data and provides insights to inform future advances in spatially resolved cell-cell interaction analysis.