Decoding Spatial Tissue Architecture: A Scalable Bayesian Topic Model for Multiplexed Imaging Analysis

Area of Science:

• Computational pathology
• Spatial biology
• Bioinformatics

Background:

• Multiplexed tissue imaging advances tumor microenvironment studies.
• Cellular neighborhood analysis faces computational and integrative challenges.
• Lack of principled strategies hinders precise spatial feature identification and tracking.

Purpose of the Study:

• Introduce SpatialTopic, a spatial topic model for high-level spatial architecture decoding.
• Integrate cell type and spatial information within a topic modeling framework.
• Overcome computational demands and improve integrative analysis across images.

Main Methods:

• Developed SpatialTopic, a spatial topic model adapting natural language processing techniques.
• Incorporated spatial information using densely overlapping image regions as documents.
• Employed an efficient collapsed Gibbs sampling algorithm for model inference.

Main Results:

• SpatialTopic demonstrates high scalability on large datasets (millions of cells).
• The model achieves high precision and interpretability in spatial feature identification.
• Consistently identifies biologically significant spatial topics like tertiary lymphoid structures (TLSs).

Conclusions:

• SpatialTopic offers computational efficiency and broad applicability across imaging platforms.
• Enables precise identification and tracking of dynamic spatial features in tissue images.
• Enhances the analysis of large-scale multiplexed tissue imaging datasets for disease research.