AI Insight
This study introduces Cellina, a computational framework that predicts how a cell's gene expression would change if its spatial neighborhood in tissue were altered. The method uses supervised disentanglement to separate a cell's intrinsic properties from its spatial context, enabling simulations of two types of interventions: rewiring connections between cells or modifying neighboring cells' expression. Testing on over 2.5 million spatially-resolved cells from colorectal cancer and mouse brain tissues, Cellina outperformed existing methods in accuracy, scalability, and ability to identify biologically meaningful cancer subregions.
Why it matters
Understanding how cells respond to changes in their spatial environment is crucial for predicting tissue behavior in disease and development. This framework could aid in designing therapeutic interventions by simulating how altering the cellular neighborhood might influence cell states in cancer and other tissue-based diseases.
arXiv:2606.08493v2 Announce Type: replace
Abstract: Tissue graph counterfactuals ask how a cell’s expression would change under altered spatial neighbor contexts. Such queries are central to predicting cell behavior in tissues, but lack a unified definition, with existing methods targeting specific intervention types or treating cells as i.i.d. In this work, we first formalize tissue graph counterfactuals as a class of spatial interventions that either rewire connections between cells (edge perturbation) or modify the expression of their neighbors (node perturbation). We then introduce Cellina (https://cellina.readthedocs.io) – a framework that uses supervised disentanglement to decompose a cell’s intrinsic state from its spatial context, using the latter as a conditioning input for counterfactual predictions. Across benchmarks spanning over 2.5 million spatially-resolved cells in colorectal cancer and mouse brain, Cellina outperforms spatially-informed and non-spatial competitors in in-silico graph perturbations, disentanglement, and scalability. Additionally, we show that Cellina reveals biologically distinct cancer subdomains in an unsupervised manner and enables targeted neighbor perturbation simulations.
Source: Querying Counterfactuals on Tissue Graphs with Supervised Disentanglement