CIRCA learns identifiable cell- and niche-level programs and their interactions with tissue pathology from spatial transcriptomics

Interactions among cells and their local microenvironments shape tissue organization and function, while their dysregulation can contribute to disease. In Alzheimer's disease (AD), the complex responses of glial cells and neurons to amyloid-β pathology can both mitigate and exacerbate neuronal injury. Spatial transcriptomics enables investigation of these relationships in situ , but recovering interpretable transcriptional programs and their interactions from spatial transcriptomics remains challenging, in part because conventional nonlinear models are non-identifiable. Here we present CIRCA (Causal Interaction Representation for Cellular Architectures), a self-supervised causal representation learning framework that jointly learns identifiable cell- and neighborhood-level transcriptional programs and the directed interactions among them under explicit assumptions. It can also be extended to incorporate additional modalities such as pathology. Benchmarked against five existing methods across seven cell-resolution datasets, CIRCA achieved the highest biological accuracy in six and the highest overall score in all five multi-section datasets, while recovering higher-order anatomical organization. In an AD mouse model, incorporating amyloid-β pathology revealed four region-specific plaque niches enriched for microglia. Plaque-associated microglia shared a core disease-associated response whose strength varied by region, and AD genetic risk was preferentially enriched in hippocampal and cortical plaque-associated microglia. Reproducible latent factors linked to cell-level amyloid burden separated phagocytic from inflammatory microglial programs. Applied to a human colorectal cancer dataset, CIRCA resolved patient-specific tumor niches, conserved stromal remodeling niches, and distinct macrophage states, highlighting its versatility. Together, these results establish CIRCA as a methodological advance over existing approaches, enabling deeper insights into tissue organization and the molecular dependencies linking cells to their local environments.

Authors

Institutions

Publication Details

Journal
bioRxiv (Cold Spring Harbor Laboratory)
Published
2026-11-18
DOI
https://doi.org/10.64898/2026.11.17.767388
Primary Topic
Single-cell and spatial transcriptomics
Type
preprint

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

CIRCA learns identifiable cell- and niche-level programs and their interactions with tissue pathology from spatial transcriptomics

Nitesh Enduru, Zhongming Zhao, Chia-Hao Tung, Tyler C. Lovelace
4 links
bioRxiv (Cold Spring Harbor Laboratory)
Single-cell and spatial transcriptomics
preprint

CIRCA learns identifiable cell- and niche-level programs and their interactions with tissue pathology from spatial transcriptomics

Nitesh Enduru, Zhongming Zhao, Chia-Hao Tung, Tyler C. Lovelace
preprint en

Abstract

Interactions among cells and their local microenvironments shape tissue organization and function, while their dysregulation can contribute to disease. In Alzheimer's disease (AD), the complex responses of glial cells and neurons to amyloid-β pathology can both mitigate and exacerbate neuronal injury. Spatial transcriptomics enables investigation of these relationships in situ , but recovering interpretable transcriptional programs and their interactions from spatial transcriptomics remains challenging, in part because conventional nonlinear models are non-identifiable. Here we present CIRCA (Causal Interaction Representation for Cellular Architectures), a self-supervised causal representation learning framework that jointly learns identifiable cell- and neighborhood-level transcriptional programs and the directed interactions among them under explicit assumptions. It can also be extended to incorporate additional modalities such as pathology. Benchmarked against five existing methods across seven cell-resolution datasets, CIRCA achieved the highest biological accuracy in six and the highest overall score in all five multi-section datasets, while recovering higher-order anatomical organization. In an AD mouse model, incorporating amyloid-β pathology revealed four region-specific plaque niches enriched for microglia. Plaque-associated microglia shared a core disease-associated response whose strength varied by region, and AD genetic risk was preferentially enriched in hippocampal and cortical plaque-associated microglia. Reproducible latent factors linked to cell-level amyloid burden separated phagocytic from inflammatory microglial programs. Applied to a human colorectal cancer dataset, CIRCA resolved patient-specific tumor niches, conserved stromal remodeling niches, and distinct macrophage states, highlighting its versatility. Together, these results establish CIRCA as a methodological advance over existing approaches, enabling deeper insights into tissue organization and the molecular dependencies linking cells to their local environments.

bioRxiv (Cold Spring Harbor Laboratory)
The University of Texas Health Science Center at Houston (US)
Cancer Prevention and Research Institute of Texas, University of Texas Health Science Center at Houston, National Institutes of Health
Good health and well-being
Single-cell and spatial transcriptomics
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon