An interpretable and biologically grounded mapping of tumor architecture through 10x Visium spatial transcriptomics
This repository presents an interpretable spatial transcriptomics analysis of primary breast tumors and matched lymph node metastases (LNMT), designed to uncover how tumor, immune, and stromal programs reorganize across disease progression. By integrating 10x Visium spatial gene expression with scRNA-seq–derived functional signatures, the analysis identifies immune exclusion zones, stromal barriers, and proliferative niches within the tumor microenvironment. It serves as both a biological case study of metastatic architecture and a methodological framework for transparent, rule-based spatial mapping that connects single-cell information to tissue-level organization.
- Spatial Logic of Metastasis: Quantifies how immune, stromal, and proliferative zones reorganize between primary tumors and lymph node metastases
- Transparent Spatial Annotation: Uses biologically grounded, rule-based region labeling instead of clustering or latent embeddings
- Multi-Scale Integration: Connects single-cell–derived gene programs to tissue-level spatial structure, linking molecular activity with morphology
The notebook follows a narrative, fully commented structure where code, figures, and interpretation are interleaved. Each section explains both the computational rationale and biological reasoning. All figures and outputs are embedded inline for an end-to-end, self-contained analysis.
📄 A static PDF version of the notebook, including all code and figures, is available in
docs/spatial_mapping_full.pdf.
📘 Interactive notebook:
spatial_mapping.ipynb
📄 Detailed documentation:
See the long README in this repository for a full description of datasets, methods, and results.
If you use or adapt this framework, please cite:
Yepes, S. Spatial Mapping of Tumor Architecture through 10x Visium Transcriptomics.
GitHub Repository, 2025. DOI: 10.5281/zenodo.17272950