Dissecting multilayer cell-cell communications with signaling feedback loops from spatial transcriptomics data

The emergence of spatial transcriptomics (ST) provides unprecedented opportunities to better decipher cell-cell communication (CCC). How to integrate spatial information and complex signaling mechanisms to infer functional CCC, however, remains a major challenge. Here, we present stMLnet, a method that takes into account spatial information and multilayer signaling regulations to identify signaling feedback loops within multilayer CCCs from ST data. To this end, stMLnet quantifies spatially dependent ligand-receptor signaling activity based on diffusion and mass action models and maps it to intracellular targets. We benchmark stMLnet against seven representative existing methods and found that stMLnet performs better in both intercellular ligand-receptor inference and intracellular target genes prediction. We apply stMLnet to analyze data from diverse spatial transcriptomics techniques like seqFISH+, Slide-seq v2, MERFISH, and Stereo-seq, verifying its robustness and scalability on ST data with varying spatial resolutions and gene coverages. Particularly, stMLnet reveals multilayer signaling feedback loops underlying the inflammatory response in ST data of COVID-19-infected lung tissue. Our study provides an effective tool for dissecting multilayer ligand/receptor-target regulations and multicellular signaling circuits from ST data, which can advance understanding of the mechanistic and functional roles of spatial CCC.