Related Experiment Video
Updated: May 11, 2026
09:19
Spatial Profiling of Protein and RNA Expression in Tissue: An Approach to Fine-Tune Virtual Microdissection
Published on: July 6, 2022
AI-predicted spatial transcriptomics unlocks breast cancer biomarkers from pathology.
Eldad D Shulman1, Emma M Campagnolo2, Roshan Lodha1
1Cancer Data Science Laboratory, Center for Cancer Research, National Cancer Institute, Bethesda, MD 20892, USA.
Cell
|May 9, 2026
Summary
Path2Space, a deep learning model, predicts spatial gene expression from histopathology slides. This cost-effective tool aids in discovering breast cancer biomarkers and predicting treatment response.
Area of Science:
- Computational biology
- Cancer research
- Genomics
Background:
- Spatial transcriptomics (ST) is crucial for understanding tumor heterogeneity.
- High costs of ST assays limit large-scale biomarker discovery.
- Need for cost-effective methods to analyze spatial gene expression in tumors.
Purpose of the Study:
- To develop a deep-learning model, Path2Space, for predicting spatial gene expression from histopathology slides.
- To assess Path2Space's performance against existing methods.
- To apply Path2Space for tumor microenvironment (TME) analysis and identify novel breast cancer subgroups.
Main Methods:
- Developed Path2Space, a deep-learning model trained on breast cancer ST data.
- Predicted spatial gene expression and inferred cell-type abundances from histopathology images.
- Analyzed TME landscapes in 976 breast cancer TCGA tumors.
- Compared Path2Space predictions with conventional bulk-sequencing biomarkers.
Main Results:
- Path2Space accurately predicted spatial gene expression, outperforming 21 established methods.
- Identified three spatially defined breast cancer subgroups with distinct survival outcomes.
- Achieved more accurate predictions of patient response to chemotherapy and trastuzumab compared to bulk-sequencing biomarkers.
- Demonstrated Path2Space's ability to map TME and infer cell-type abundances.
Conclusions:
- Path2Space provides a scalable, fast, and cost-effective alternative to molecular assays for spatial gene expression analysis.
- Enables large-cohort treatment biomarker discovery and provides translational insights into tumor biology.
- Potential applicability across various cancer indications for biomarker discovery and TME characterization.
