Node-centric expression models (NCEM): graphical neural networks reveal communication between cells

Modeling ligand-receptor signaling with NCEM. (a) UMAP of cells in MERFISH-fetal liver data. (b) Imputation of MERFISH data with scRNA-seq increases the number of genes that can be modeled with NCEMs, including receptor genes, ligand genes, and ligand-receptor pairs. (c) Distribution of selected marker genes that are both observed in scRNA-seq and in MERFISH on cell types. Credit: Natural biotechnology (2022). DOI: 10.1038/s41587-022-01467-z

Cells interact in different ways and on several length scales. The interaction of a cell with its tissue niche can be described through cellular communication events. To understand these events, researchers around the world create models, based on different strategies. This knowledge is crucial for understanding and identifying emerging phenomena in tissue microenvironments, such as genetic changes in a tumor.

The problem is that many models are based on dissociated cells, which means that the cells are separated into individual cells during the analysis and are no longer integrated into their natural environment. Other models are limited to receptor-ligand signaling, a certain type of communication between cells.

These models therefore ignore the spatial proximity of a group of cells (a niche) in their natural tissue environment. Researchers led by Fabian Theis from the Computational Health Center Helmholtz Munich and the Technical University of Munich (TUM) have now developed a new method that defines the complexity and improves the understanding of cellular communication: expression-centered patterns. nodes (NCEM).

A flexible framework

NCEM is a computational method based on graph neural networks, which combines transcriptomic variance attribution and cellular communication modeling in a unique model of tissue niches.

The model is therefore able to predict the gene expression profile of a cell based on the presence of surrounding cell types. Furthermore, it estimates the effect of tissue niche composition on gene expression in an unbiased manner from spatial molecular profiling data.

In their model, the researchers developed a flexible framework to explain gene expression variations observable in spatial transcriptomics, a technology providing spatially resolved gene expression information. Variations in gene expression can then be associated with known molecular processes associated with cellular communication events.

They showed that NCEMs robustly identify cell-cell dependencies across different spatial transcriptomics technologies and at length scales characteristic of known communication mechanisms. With this method, the first authors David Fischer and Anna Schaar were able to recover signatures of molecular processes known to underlie cellular communication.

A new way to identify cellular communication

The framework constrains communication events to cells that are close in space. The dependencies identified are not limited to ligand-receptor based communication but can also explain, for example, physical interactions or metabolite exchanges.

NCEM is a flexible computational method that can be extended to more complex datasets, such as 3D spatial transcriptomics data and high-throughput data. It therefore provides a set of flexible tools for the analysis of cell-cell communication in space. The new methodology complements recent efforts on the characterization of gene expression in single cells in single-cell “atlas” projects by taking into account in this particular case the tissue niche.

The research is published in Natural biotechnology.