A Counterfactual Framework for Directional Cell-Cell Interaction Analysis in Spatial Transcriptomics.

Area of Science:

• Spatial transcriptomics
• Cell-cell communication analysis
• Computational biology

Background:

• Current spatial transcriptomics methods often lack explicit directionality testing for cell-cell interactions.
• Reliance on correlation or predefined ligand-receptor pairs limits understanding of causal cellular influence.
• Need for robust methods to infer directional cell-cell communication.

Purpose of the Study:

• To develop a novel framework for inferring directional cell-cell influence in spatial transcriptomics.
• To introduce a ligand-receptor-agnostic approach that tests sender specificity.
• To quantify directional communication using a Counterfactual Directionality Score (CDS).

Main Methods:

• Developed a neighborhood-conditioned graph model to predict receiver cell states.
• Employed a counterfactual intervention strategy by replacing neighboring cells to measure state changes.
• Calculated pair-level CDS by aggregating results across receiver cells and test cores.

Main Results:

• Identified reproducible, asymmetric interactions in cholangiocarcinoma tissue microarrays.
• Prominent directional influences include Tumor-EMT→Macrophage (CDS=0.0828) and Fibroblast→Macrophage (CDS=0.0582).
• Results were statistically significant (p<0.001) and robust against null models and resampling, correlating well with ligand-receptor scores (r=0.758).

Conclusions:

• Counterfactual testing provides a statistically rigorous and scalable method for directional cell-cell communication analysis.
• The framework successfully infers specific, directional interactions in complex tissue microenvironments.
• This approach enhances the understanding of cellular crosstalk in spatial transcriptomics.