Detecting clinically relevant topological structures in multiplexed spatial proteomics using TopKAT

Area of Science:

• Computational biology
• Cancer research
• Statistical methods

Background:

• Multiplexed spatial proteomics reveals complex cell arrangements within the tumor microenvironment (TME).
• Cellular spatial organization in the TME is clinically significant, impacting disease prognosis and treatment efficacy.
• Existing statistical methods are insufficient for analyzing cell-level spatial data against patient-level outcomes.

Purpose of the Study:

• To introduce a novel statistical framework, the topological kernel association test (TopKAT), for analyzing spatial proteomics data.
• To assess the association between geometric cell structures in the TME and patient outcomes (continuous, binary, survival).
• To provide a robust method for predicting clinical outcomes based on spatial cell topology.

Main Methods:

• TopKAT integrates persistent homology to quantify topological features of cell arrangements using persistence diagrams.
• It employs kernel testing to compare similarities between persistence diagrams based on homology count and lifespan.
• The method analyzes cell-level spatial data to identify associations with patient-level clinical outcomes.

Main Results:

• TopKAT demonstrates potential for increased statistical power compared to existing methods, especially for ring-like cellular structures.
• The study validates TopKAT's utility in analyzing breast and colorectal cancer datasets.
• The method effectively links spatial cell topology to patient outcomes.

Conclusions:

• TopKAT offers a powerful new approach for analyzing complex spatial proteomics data in cancer research.
• This method can improve the prediction of patient prognosis and treatment response by incorporating spatial cell information.
• TopKAT provides a valuable tool for understanding the clinical implications of cellular geometry in the TME.