Defining cancer spatial ecotypes

2026-06-11 · Source: Machine learning : nature.com subject feeds · Field: Science & Research — Life Sciences & Biology, Health & Medical Research, Mathematics & Computational Sciences · Depth: Expert, quick

Summary

Aaron Newman at Stanford University, Aadel Chaudhuri at Mayo Clinic, and their colleagues have developed new machine learning methods to dissect the tumor microenvironment (TME). Their method, called Spatial EcoTyper, utilizes a large collection of spatial transcriptomics data from various cancer types. It adapts a similarity network fusion approach to identify shared multicellular patterns in a common embedding across samples. Non-negative matrix factorization then revealed nine robust spatial clusters, termed "spatial ecotypes" (SEs). These SEs were validated in independent datasets from different technical platforms and showed associations with specific spatial patterns, molecular, and clinical features. A key finding was the existence of SE-specific expression signatures, largely agnostic to cell type, distinguishing SEs by acute stress response, wound healing, immunosuppression, neovascularization, and interferon signaling, among other biological hallmarks of the TME, as published in Nat Methods 23, 1074 (2026).

Key takeaway

For research scientists analyzing tumor microenvironments, this new Spatial EcoTyper method offers a robust approach to identify distinct spatial ecotypes and their associated molecular features. You should consider integrating this machine learning framework to uncover cell-type-agnostic expression signatures, potentially revealing novel therapeutic targets or prognostic markers in cancer research. This can enhance understanding of complex TME interactions and guide more precise therapeutic strategies.

Key insights

Spatial EcoTyper uses ML to define nine robust spatial ecotypes in the tumor microenvironment, revealing cell-type-agnostic expression signatures.

Principles

• Cancer progression involves complex multicellular and multimodal interactions.
• Spatial transcriptomics data can reveal shared multicellular patterns.
• Tumor microenvironment features include immunosuppression and neovascularization.

Method

Spatial EcoTyper adapts similarity network fusion and non-negative matrix factorization to identify shared multicellular patterns and nine robust spatial ecotypes from spatial transcriptomics data.

In practice

• Apply similarity network fusion for common embedding.
• Use non-negative matrix factorization for cluster identification.
• Validate findings across independent datasets.

Topics

• Machine Learning
• Tumor Microenvironment
• Spatial Transcriptomics
• Cancer Research
• Spatial EcoTyper
• Non-negative Matrix Factorization

Best for: AI Scientist, Research Scientist

Related on AIssential

• Non-invasive profiling of the tumour microenvironment with spatial ecotypes — A new machine learning framework enables non-invasive profiling of tumor microenvironment spatial ecotypes from liquid biopsies, predicting immunotherapy response.
• Systematically decoding pathological morphologies and molecular profiles with unified multimodal embedding — Multi-Embed unifies morphological and molecular data for superior disease understanding and prediction.
• GigaTIME: Scaling tumor microenvironment modeling using virtual population generated by multimodal AI — GigaTIME uses multimodal AI to convert common H&E slides into virtual mIF images, enabling population-scale precision immuno-oncology research.
• Querying Counterfactuals on Tissue Graphs with Supervised Disentanglement — Cellina disentangles intrinsic cell state from spatial context to predict counterfactual cell expression on tissue graphs.
• DUET: Dual-Paradigm Adaptive Expert Triage with Single-cell Inductive Prior for Spatial Transcriptomics Prediction — DUET unifies regression and retrieval with single-cell priors for robust spatial gene expression prediction from histology.
• Beyond Morphology: Quantifying the Diagnostic Power of Color Features in Cancer Classification — Pixel intensity alone, without morphology, provides a strong diagnostic signal for cancer classification.

Open in AIssential →

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