Cellular neighborhoods – recurring, spatially confined groups of interacting cells with distinct compositions – are reshaping how scientists understand tumors. Advances in spatial omics and computational tools now let researchers map these local ecosystems in unprecedented detail, showing how they drive tumor diversity, growth, immune evasion, and treatment response. In the journal Nature Cancer, researchers at Children’s Hospital of Philadelphia (CHOP) propose a clearer framework for defining, classifying, and comparing cellular neighborhoods across cancers.
Spatial omics combines molecular profiling (genomics, transcriptomics, proteomics) with spatial context to map molecules within cells in intact tissues. It reveals how cellular interactions and microenvironments drive disease progression and treatment response, supporting precise diagnostics and targeted therapies. Key early steps in analysis are cell segmentation (identifying cell boundaries) and classification (assigning cell types).
Recent advances enable cellular-level analysis of intact tissues, linking architecture to function. CHOP collaborates on major projects such as the Human Tumor Atlas Network, HuBMAP, and the BRAIN Initiative, using these methods to map healthy and diseased tissues.
In this newly published perspective article, researchers reviewed the newest imaging and sequencing methods that measure RNA, proteins, and metabolites while keeping their location in the tissue. They also reviewed computer tools that find groups of nearby cells based on cell types, molecular signals and how close cells are to each other.
The researchers proposed that cellular neighborhoods have five main traits. They recommend developing a standard approach to compare methods across different platforms. These traits include a distinctive mix of cell types, a limited physical area, repeated occurrence, stronger interactions between the cells and a clear role in the tissue. Cellular neighborhoods can also be described by what they’re made of (for example, cancer cells, immune cells, or support cells), what they do (such as growing, being low on oxygen, suppressing the immune system, supporting stem cells, or sharing metabolic tasks), and where they’re located (like the center, the border, or next to non-tumor tissue).
“Increasing evidence suggests neighborhood-level measures often predict clinical outcomes and therapy responses better than nonspatial metrics,” said senior author Kai Tan, PhD, a professor in the Department of Pediatrics at Children’s Hospital of Philadelphia who spearheads CHOP’s participation in the National Cancer Institute (NCI) Human Tumor Atlas Network (HTAN). “For example, immune-rich neighborhoods often predict better responses to immunotherapy, while neighborhoods dominated by tumor cells and fibroblasts often indicate a worse outcome.”
Tan and colleagues urge larger, more diverse spatial-omics datasets and shared public resources so researchers can agree on standard neighborhood definitions and fairly compare computational tools. They also call for experiments that combine spatial profiling with targeted genetic changes and human-relevant lab models, moving studies from maps and descriptions to tests that show which neighborhoods drive cancer behavior.
The authors anticipate that AI will make it easier and faster to find cell neighborhoods, combine different types of molecular data and help turn this knowledge into precision tests and treatments that target specific areas in tissue.
This work was supported by grants from the National Institutes of Health (NIH) under award numbers U2C CA233285 and U54HL165442.
Ma et al. “Cellular Neighborhoods in Cancer.” Nature Cancer. Online January 16, 2026, 2026. DOI:10.1038/s43018-025-01107-w.
Featured in this article
Experts
Research
Cellular neighborhoods – recurring, spatially confined groups of interacting cells with distinct compositions – are reshaping how scientists understand tumors. Advances in spatial omics and computational tools now let researchers map these local ecosystems in unprecedented detail, showing how they drive tumor diversity, growth, immune evasion, and treatment response. In the journal Nature Cancer, researchers at Children’s Hospital of Philadelphia (CHOP) propose a clearer framework for defining, classifying, and comparing cellular neighborhoods across cancers.
Spatial omics combines molecular profiling (genomics, transcriptomics, proteomics) with spatial context to map molecules within cells in intact tissues. It reveals how cellular interactions and microenvironments drive disease progression and treatment response, supporting precise diagnostics and targeted therapies. Key early steps in analysis are cell segmentation (identifying cell boundaries) and classification (assigning cell types).
Recent advances enable cellular-level analysis of intact tissues, linking architecture to function. CHOP collaborates on major projects such as the Human Tumor Atlas Network, HuBMAP, and the BRAIN Initiative, using these methods to map healthy and diseased tissues.
In this newly published perspective article, researchers reviewed the newest imaging and sequencing methods that measure RNA, proteins, and metabolites while keeping their location in the tissue. They also reviewed computer tools that find groups of nearby cells based on cell types, molecular signals and how close cells are to each other.
The researchers proposed that cellular neighborhoods have five main traits. They recommend developing a standard approach to compare methods across different platforms. These traits include a distinctive mix of cell types, a limited physical area, repeated occurrence, stronger interactions between the cells and a clear role in the tissue. Cellular neighborhoods can also be described by what they’re made of (for example, cancer cells, immune cells, or support cells), what they do (such as growing, being low on oxygen, suppressing the immune system, supporting stem cells, or sharing metabolic tasks), and where they’re located (like the center, the border, or next to non-tumor tissue).
“Increasing evidence suggests neighborhood-level measures often predict clinical outcomes and therapy responses better than nonspatial metrics,” said senior author Kai Tan, PhD, a professor in the Department of Pediatrics at Children’s Hospital of Philadelphia who spearheads CHOP’s participation in the National Cancer Institute (NCI) Human Tumor Atlas Network (HTAN). “For example, immune-rich neighborhoods often predict better responses to immunotherapy, while neighborhoods dominated by tumor cells and fibroblasts often indicate a worse outcome.”
Tan and colleagues urge larger, more diverse spatial-omics datasets and shared public resources so researchers can agree on standard neighborhood definitions and fairly compare computational tools. They also call for experiments that combine spatial profiling with targeted genetic changes and human-relevant lab models, moving studies from maps and descriptions to tests that show which neighborhoods drive cancer behavior.
The authors anticipate that AI will make it easier and faster to find cell neighborhoods, combine different types of molecular data and help turn this knowledge into precision tests and treatments that target specific areas in tissue.
This work was supported by grants from the National Institutes of Health (NIH) under award numbers U2C CA233285 and U54HL165442.
Ma et al. “Cellular Neighborhoods in Cancer.” Nature Cancer. Online January 16, 2026, 2026. DOI:10.1038/s43018-025-01107-w.
Recommended reading
CHOP Researchers Develop Cell Atlas to Uncover Significant Shifts in the Neuroblastoma Tumor Microenvironment
CHOP researchers developed a longitudinal atlas of neuroblastoma, a common and potentially deadly childhood cancer, to gain deeper understanding into precise molecular mechanisms underlying why and how certain treatments eventually become ineffective.
CHOP Researchers Announce New AI Model for Cell Segmentation and Classification
Researchers at CHOP announced the creation of a new AI technology called CelloType, a comprehensive model designed to more accurately identify and classify cells in high-content tissue images.
Contact us
Karen Pasternack