Counterfactual Diffusion Modeling Enables Spatially Targeted Reprogramming of Tissue Microenvironments
Wenhui Ding1, Zhenhua Luo1, Yuanyan Xiong2
1Institute of Precision Medicine, The First Affiliated Hospital, Sun Yat-sen University, Guangzhou 510080, China.
Biology
|July 28, 2026
Summary
This study introduces SPAD-CFR, a new computational model for spatial single-cell analysis. It simulates how tissues change after perturbations, integrating molecular data with spatial structure for causal inference.
Area of Science:
- Computational Biology
- Systems Biology
- Genomics
Background:
- Spatially resolved single-cell technologies offer insights into tissue heterogeneity but lack mechanistic understanding of perturbations.
- Existing computational models often neglect spatial information or tissue geometry, limiting their ability to study topological remodeling.
- There is a need for models that integrate molecular profiles with dynamic spatial structures to understand tissue responses.
Purpose of the Study:
- To develop a novel computational framework, SPAD-CFR (Spatial Point-cloud Attention-based Diffusion for CounterFactual Reprogramming), for spatial single-cell data analysis.
- To enable causal inference and counterfactual simulations of tissue responses to perturbations by integrating molecular and spatial information.
- To validate the model's ability to reproduce biological structures and simulate cellular reprogramming in various tissue contexts.
Main Methods:
- Representing tissues as spatial point clouds with cellular molecular profiles and physical coordinates.
- Implementing Pearl's three-step causal inference workflow using deterministic diffusion inversion and sampling.
- Applying interventions to individual cells to generate counterfactual tissues with altered molecular profiles and spatial coordinates.
Main Results:
- SPAD-CFR successfully reproduced the hierarchical structure of the mouse cerebral cortex.
- The model simulated phenotypic differences across breast cancer histological grades and reconstructed hypoxia-associated phenotypes in triple-negative tumors.
- Simulations in melanoma demonstrated distance-dependent bystander cytotoxic effects following T cell activation interventions.
Conclusions:
- SPAD-CFR provides a biologically informed generative framework for spatial simulations.
- The model facilitates counterfactual analysis to validate hypotheses regarding microenvironment reprogramming.
- This approach advances the understanding of tissue dynamics and cellular responses in complex biological systems.
