ISPAT-3D: Spatially Varying Conditional Volumetric Network Estimation for 3D Tumor Imaging

Area of Science:

• Computational Biology
• Cancer Research
• Bioinformatics

Background:

• The tumor microenvironment's spatial organization impacts immune function and disease progression.
• Current methods for analyzing cell interactions in tissues are limited to 2D and overlook spatial auto-correlation.
• Understanding 3D spatial interactions is vital for comprehensive cancer analysis.

Purpose of the Study:

• To introduce ISPat-3D, a novel hierarchical Bayesian framework for analyzing 3D multiplexed cancer imaging data.
• To recover spatially varying, zone-specific cell-type interaction networks from 3D tissue volumes.
• To identify volumetric spatial conditional interactions not detectable in 2D sections.

Main Methods:

• ISPat-3D partitions tissue into tumor intensity zones.
• It employs anisotropic Gaussian processes with zone-specific lengthscales for cell-type modeling.
• Residual decomposition via multi-study factor analysis and partial correlation network extraction from precision matrices are utilized.

Main Results:

• Simulations confirm accurate recovery of shared and zone-specific structures with high power and controlled false discovery rate (FDR).
• Application to colorectal cancer (CRC1) data shows T cell module intensification with tumor burden and shifting CD4+/CD8+ T cell regulatory associations.
• Analysis of breast carcinoma (BC) data reveals conditional coupling between cancer-associated fibroblasts (CAFs) and the myoepithelial layer, along with zone-specific CAF-endothelial and B cell-CAF interactions.

Conclusions:

• ISPat-3D effectively identifies 3D spatial cell-type interactions within distinct tumor zones.
• The framework provides insights into immune suppression mechanisms and tumor progression dynamics.
• This 3D approach uncovers critical spatial relationships missed by 2D analyses, advancing cancer imaging interpretation.