SpatialProp: tissue perturbation modeling with spatially resolved single-cell transcriptomics.

Area of Science:

• Computational Biology
• Systems Biology
• Genomics

Background:

• Perturbational studies are crucial for understanding biological causality.
• Current methods like Perturb-seq analyze single cells but not intact tissues.
• Predicting genetic perturbation effects in complex tissue microenvironments remains a challenge.

Purpose of the Study:

• To develop a computational framework for predicting multi-gene, multi-cell type perturbations in whole tissue sections.
• To leverage spatially resolved transcriptomics data for training predictive models.
• To enable in silico perturbation experiments for studying spatially patterned tissue biology.

Main Methods:

• Developed SpatialProp, a graph neural network-based computational framework.
• Utilized spatially resolved single-cell transcriptomics datasets to train SpatialProp.
• Introduced CausalInteractionBench for benchmarking causal enrichment in spatial perturbation predictions.

Main Results:

• SpatialProp can predict gene expression based on the tissue microenvironment.
• The framework successfully maps tissue microenvironments to new target states.
• Evaluated the causal utility of SpatialProp in predicting spatial perturbation effects using CausalInteractionBench.

Conclusions:

• SpatialProp offers a novel approach to study the effects of genetic perturbations in intact tissues.
• The framework facilitates rapid hypothesis generation and in silico experiments.
• SpatialProp is particularly valuable for investigating spatially patterned tissue biology.