SpatioCAD: Context-aware graph diffusion model for pinpointing spatially variable genes in heterogeneous tissues
- 1School of Mathematical Sciences, Fudan University, Shanghai, China
- 2Department of Cardiology, First Affiliated Hospital of Ningbo University (Ningbo First Hospital), School of Medicine, Ningbo University, Ningbo, China
- 3Center for Applied Mathematics, Center for Integrative Spatial-Omics Research, and Shanghai Key Laboratory for Contemporary Applied Mathematics, Fudan University, Shanghai, China
Abstract
Spatial transcriptomics enables comprehensive characterization of tissue architecture, and the identification of spatially variable genes (SVGs) is a critical step for defining region-specific molecular markers and uncovering spatially regulated mechanisms across diverse biological contexts. However, most existing methods for SVG detection overlook cell density variations, a major confounding factor in complex tissues such as tumors, where heterogeneous cellularity frequently introduces false-positive calls. Here we present SpatioCAD, a computational framework that explicitly decouples genuine spatial expression patterns from confounding effects driven by cellularity. SpatioCAD introduces a graph diffusion model that inherently incorporates cell-density-aware conditions to simulate expression propagation, supporting the identification of SVGs across a broad range of expression levels. Systematic evaluations on simulated datasets demonstrate its superior statistical power and specificity. Applied to breast cancer, lung cancer, and glioma datasets, SpatioCAD identifies functionally diverse SVGs, including low‑abundance transcripts with established roles in tumor progression, while also recapitulates biologically meaningful tissue architecture features. Importantly, SpatioCAD is fast and scalable to large tissues. Author summary: Spatial transcriptomics maps where genes are active in tissues, providing a way to explore how tissue organization relates to biological function and disease. Interpreting these maps is challenging when different regions contain very different numbers of cells, as is common in tumors. Such differences can influence observed gene expression patterns and complicate the identification of informative spatial variation. We developed SpatioCAD to identify genes with spatially patterned expression while accounting for density-related variation. Using simulated data and datasets from human cancers and normal brain tissue, we show that SpatioCAD identifies genes with coherent spatial patterns across a broad range of expression levels. In cancer tissues, these genes represent diverse biological functions, providing a complementary view of molecular differences within the tumor environment. By supporting the detection of patterns in genes with both high and low expression, SpatioCAD helps researchers explore tissue heterogeneity more broadly. Our work provides a computational tool for investigating the molecular organization of complex tissues and generating hypotheses for further biological study.
Related
- Optimized dosing of spatially heterogeneous radiopharmaceutical therapy reprograms poorly immunogenic tumors to enhance response to immune checkpoint blockade
- Spatially heterogeneous restoration of spermatogenesis in the mouse testis following experimental cryptorchidism and orchidopexy
- Mapping Spatially Heterogeneous and History-Dependent Hydrogel Mechanics Using Fluid-Driven Cyclic Shear
- Spatially-resolved multimodal profiling identifies functionally-heterogeneous cancer-associated fibroblasts associated with poor radiotherapy outcomes in muscle-invasive bladder cancer.
- Variable Energy Availability Exposure and Heterogeneous Physiological Responses to a Prescribed Low Energy Availability Intervention in Free-Living Male Endurance Athletes: A Pilot Case-Series.
BibTeX
@article{wen2026spatiocad,
title = {{SpatioCAD: Context-aware graph diffusion model for pinpointing spatially variable genes in heterogeneous tissues}},
author = {Hao Wen and Qunlun Shen and Shuqin Zhang},
journal = {PLOS Computational Biology},
year = {2026},
month = nov,
doi = {10.1371/journal.pcbi.1014923},
url = {https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1014923},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - SpatioCAD: Context-aware graph diffusion model for pinpointing spatially variable genes in heterogeneous tissues
AU - Hao Wen
AU - Qunlun Shen
AU - Shuqin Zhang
PY - 2026
DA - 2026/11/02
JO - PLOS Computational Biology
DO - 10.1371/journal.pcbi.1014923
UR - https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1014923
AB - Spatial transcriptomics enables comprehensive characterization of tissue architecture, and the identification of spatially variable genes (SVGs) is a critical step for defining region-specific molecular markers and uncovering spatially regulated mechanisms across diverse biological contexts. However, most existing methods for SVG detection overlook cell density variations, a major confounding factor in complex tissues such as tumors, where heterogeneous cellularity frequently introduces false-positive calls. Here we present SpatioCAD, a computational framework that explicitly decouples genuine spatial expression patterns from confounding effects driven by cellularity. SpatioCAD introduces a graph diffusion model that inherently incorporates cell-density-aware conditions to simulate expression propagation, supporting the identification of SVGs across a broad range of expression levels. Systematic evaluations on simulated datasets demonstrate its superior statistical power and specificity. Applied to breast cancer, lung cancer, and glioma datasets, SpatioCAD identifies functionally diverse SVGs, including low‑abundance transcripts with established roles in tumor progression, while also recapitulates biologically meaningful tissue architecture features. Importantly, SpatioCAD is fast and scalable to large tissues. Author summary: Spatial transcriptomics maps where genes are active in tissues, providing a way to explore how tissue organization relates to biological function and disease. Interpreting these maps is challenging when different regions contain very different numbers of cells, as is common in tumors. Such differences can influence observed gene expression patterns and complicate the identification of informative spatial variation. We developed SpatioCAD to identify genes with spatially patterned expression while accounting for density-related variation. Using simulated data and datasets from human cancers and normal brain tissue, we show that SpatioCAD identifies genes with coherent spatial patterns across a broad range of expression levels. In cancer tissues, these genes represent diverse biological functions, providing a complementary view of molecular differences within the tumor environment. By supporting the detection of patterns in genes with both high and low expression, SpatioCAD helps researchers explore tissue heterogeneity more broadly. Our work provides a computational tool for investigating the molecular organization of complex tissues and generating hypotheses for further biological study.
SN - 1553-7358
ER -