Celcomen: spatial causal disentanglement for single-cell and tissue perturbation modeling

TL;DR

Celcomen introduces a graph neural network framework that disentangles gene regulation programs in spatial transcriptomics, enabling counterfactual predictions and insights into tissue responses for disease modeling.

Contribution

It presents a novel causality-based generative model for spatial transcriptomics that disentangles intra- and inter-cellular gene regulation, with validated counterfactual prediction capabilities.

Findings

Successfully disentangles gene regulation programs in spatial data

Generates accurate post-perturbation spatial transcriptomics

Provides insights into tissue responses in human diseases

Abstract

Celcomen leverages a mathematical causality framework to disentangle intra- and inter- cellular gene regulation programs in spatial transcriptomics and single-cell data through a generative graph neural network. It can learn gene-gene interactions, as well as generate post-perturbation counterfactual spatial transcriptomics, thereby offering access to experimentally inaccessible samples. We validated its disentanglement, identifiability, and counterfactual prediction capabilities through simulations and in clinically relevant human glioblastoma, human fetal spleen, and mouse lung cancer samples. Celcomen provides the means to model disease and therapy induced changes allowing for new insights into single-cell spatially resolved tissue responses relevant to human health.