Benchmarking tools for deciphering cellular crosstalk in spatially-resolved transcriptomics.

Area of Science:

• Spatial Transcriptomics
• Computational Biology
• Cell-Cell Communication

Background:

• Cell-cell communication via ligand-receptor signaling is crucial for multicellular organization.
• Spatial transcriptomics (ST) advances enable inference of interactions within native tissue contexts.
• Limited comprehensive benchmarking exists for ST-specific cell-cell interaction (CCI) inference tools.

Purpose of the Study:

• To comprehensively evaluate and benchmark computational CCI inference methods for ST applications.
• To assess tool performance across diverse simulation settings and real ST datasets.
• To provide practical guidance for selecting appropriate tools for spatially resolved CCI analysis.

Main Methods:

• Evaluated nine computational CCI inference methods.
• Utilized realistic simulation settings and nine real ST datasets from Visium, Stereo-seq, and Xenium platforms.
• Assessed performance based on prediction accuracy, spatial coherence, biological relevance, and computational efficiency.

Main Results:

• Demonstrated substantial variability in tool performance across different spatial resolutions, tissue contexts, and ST platforms.
• Identified key challenges in applying existing CCI tools to real-world ST data.
• Provided insights into ligand-receptor prediction accuracy, spatial interaction coherence, and pathway enrichment.

Conclusions:

• Tool performance is highly context-dependent, necessitating careful selection for specific ST applications.
• Highlights the need for improved computational tools tailored for the complexities of real ST data.
• Offers guidance for researchers and informs future development in spatially resolved cell-cell interaction analysis.