Computational pathology annotation enhances the resolution and interpretation of breast cancer spatial...

Area of Science:

• Oncology
• Computational Pathology
• Genomics

Background:

• Breast cancer exhibits significant intra-tumoral heterogeneity impacting patient outcomes.
• Understanding the spatial distribution of this heterogeneity is crucial for diagnosis and treatment.
• Current spatial transcriptomics tools lack single-cell resolution, limiting data interpretability.

Purpose of the Study:

• To develop and validate a computational pathology image analysis pipeline for high-resolution mapping of breast tumor microenvironments.
• To enhance the analysis of spatial transcriptomic data by integrating computational tissue annotation (CTA).
• To deepen insights into intra-tumoral heterogeneity and its spatial architecture in breast cancer.

Main Methods:

• Development of a machine learning-based computational tissue annotation (CTA) pipeline.
• Application of CTA to map tumor, stroma, and immune compartments in Visium-assayed breast tumor sections.
• Integration of CTA with spatial transcriptomic data from 23 breast tumor sections across four patients.

Main Results:

• CTA provided high-resolution annotations on H&E-stained images, correlating with spatial sequencing data.
• The pipeline improved the resolution of intra-tumoral heterogeneity analysis.
• CTA facilitated spatially resolved intrinsic subtyping and enhanced visualization of lymphocyte clones at single-cell resolution.

Conclusions:

• The CTA pipeline offers valuable insights into the spatial architecture of breast cancer.
• This approach enhances the interpretability of spatial transcriptomics data.
• The findings contribute to more personalized diagnostics and treatment strategies for breast cancer patients.