1176-CytoCommunity: Decoding Tissue Cellular Neighborhoods

The paper introduce CytoCommunity, a novel graph neural network algorithm designed to identify tissue cellular neighborhoods (TCNs) from spatial omics data. By leveraging both unsupervised and supervised learning, the tool maps cell phenotypes and their physical distributions to discover functional spatial domains without requiring intermediate clustering. Researchers demonstrated its superior performance over existing methods in accurately defining tissue compartments across diverse technologies, such as CODEX and MERFISH. In clinical applications, the algorithm successfully identified condition-specific TCNs that distinguish high-risk from low-risk tumors in breast and colorectal cancers. These findings highlight how specific cell-cell communication patterns, such as interactions between neoplastic cells and fibroblasts, contribute to disease progression and patient prognosis. Ultimately, CytoCommunity provides a robust framework for uncovering the complex structural-functional relationships within various tissue microenvironments.

References:

• Hu Y, Rong J, Xu Y, Xie R, Peng J, Gao L, Tan K. Unsupervised and supervised discovery of tissue cellular neighborhoods from cell phenotypes. Nat Methods. 2024 Feb;21(2):267-278. doi: 10.1038/s41592-023-02124-2. Epub 2024 Jan 8. PMID: 38191930; PMCID: PMC10864185.