CellNiche represents cellular microenvironments in atlas-scale spatial omics data with contrastive learning

2026-04-22 · Source: Machine learning : nature.com subject feeds · Field: Science & Research — Life Sciences & Biology, Health & Medical Research, Mathematics & Computational Sciences · Depth: Expert, long

Summary

CellNiche is a scalable contrastive-learning framework designed to identify and characterize cellular microenvironments from atlas-scale spatial omics data. Published on April 22, 2026, in Nature Communications, this framework utilizes cell-centric spatial-proximity subgraphs and combines spatial co-localization with molecular co-expression cues to generate microenvironment-aware embeddings. Tested across multiple spatial omics datasets, totaling over 10 million cells, CellNiche demonstrated improved representations with increased training data and competitive performance in clustering and embedding quality, all while maintaining computational efficiency. The framework successfully identified conserved and sample-specific tumor and immune microenvironments in a multi-sample human non-small-cell lung cancer (NSCLC) cohort, capturing localized spatial transitions. Furthermore, it integrated 293 slices from four independent mouse brain atlases into a unified virtual brain map, enabling cross-atlas annotation transfer and spatial refinement.

Key takeaway

For computational biologists and cancer researchers analyzing spatial omics data, CellNiche offers a robust method to decipher complex cellular microenvironments. Its ability to integrate diverse datasets and identify both conserved and sample-specific features can significantly enhance your understanding of disease progression and tissue organization. Consider applying CellNiche to large-scale spatial omics projects to improve data integration and microenvironment characterization.

Key insights

CellNiche uses contrastive learning on spatial omics data to map cellular microenvironments across diverse biological contexts.

Principles

• Combine spatial and molecular cues for robust embeddings.
• Leverage contrastive learning for scalable representation.
• Integrate heterogeneous datasets for unified atlases.

Method

CellNiche employs a contrastive-learning framework with cell-centric spatial-proximity subgraphs, integrating spatial co-localization and molecular co-expression to learn microenvironment-aware embeddings from spatial omics data.

In practice

• Analyze tumor microenvironments in NSCLC cohorts.
• Integrate multiple spatial omics datasets for unified maps.
• Transfer annotations across different brain atlases.

Topics

• Cellular Microenvironments
• Spatial Omics Data
• Contrastive Learning
• CellNiche Framework
• Non-Small-Cell Lung Cancer

Code references

• drieslab/spatial-datasets
• Super-LzzZ/CellNiche

Best for: AI Scientist, Research Scientist

Related on AIssential

• Machine learning framework reveals a concordant cell-state landscape across single-cell datasets — CONCORD is a contrastive learning framework for integrating single-cell omics data to reveal cell-state landscapes.
• Whole mouse mapping — New spatial genomics tools enable whole-body transcriptomics and high-resolution cell type identification in mice.
• Latent space mapping of interpretable structural coordinates from stochastic single-molecule signals — Latent-space mapping resolves stochastic signal warping in nanopore sensors, enabling interpretable molecular coordinate systems.
• Conditional Monge Gap enables generalizable single-cell perturbation modelling — CMonge is a conditional optimal transport framework that accurately predicts single-cell perturbation responses and generalizes to unseen conditions.
• Querying Counterfactuals on Tissue Graphs with Supervised Disentanglement — Cellina disentangles intrinsic cell state from spatial context to predict counterfactual cell expression on tissue graphs.
• AI to help researchers see the bigger picture in cell biology — An AI framework disentangles shared and unique cellular data across measurement modalities for a holistic cell view.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine learning : nature.com subject feeds.