Detecting Clinically Relevant Topological Structures in Multiplexed Spatial Proteomics Imaging Using TopKAT.

Area of Science:

• Computational Biology and Bioinformatics
• Cancer Research and Oncology
• Statistical Modeling and Data Analysis

Background:

• Spatial proteomics imaging reveals cell architecture within the tumor microenvironment (TME).
• Cellular spatial context in the TME impacts disease prognosis and treatment response.
• Existing statistical methods struggle to robustly analyze spatial cell geometry for clinical endpoint association.

Purpose of the Study:

• To develop novel statistical models for associating cell-level spatial images with patient-level clinical endpoints.
• To introduce a topology-based method capable of characterizing spatial cell arrangements and testing clinical associations.
• To evaluate the performance of the proposed method against existing spatial statistical tests.

Main Methods:

• Proposed a topology-based approach combining persistent homology with kernel testing, termed TopKAT (Topological Kernel Association Test).
• Utilized TopKAT to assess the predictive power of cellular topological structures on continuous, binary, and survival clinical endpoints.
• Conducted simulation studies to demonstrate TopKAT's statistical properties and power.

Main Results:

• TopKAT demonstrated superior power compared to spatial point process models, especially for ring-like cellular structures.
• The method successfully identified clinically relevant topological patterns in the spatial distribution of immune and tumor cells.
• Application to triple-negative breast cancer datasets validated TopKAT's ability to recover meaningful spatial structures.

Conclusions:

• TopKAT offers a powerful and robust framework for analyzing spatial cell distributions in the TME.
• The topological approach effectively links cellular spatial architecture to patient-level clinical outcomes.
• This method advances the statistical analysis of spatial omics data in cancer research.