Topology across scales on heterogeneous cell data

Area of Science:

• Computational biology
• Topological data analysis
• Bioinformatics

Background:

• Multiplexed imaging generates large, spatially-resolved datasets of multiple cell types within tissues.
• Persistent homology (PH) offers multiscale shape descriptors valuable for analyzing spatial biological data.
• Current vectorization methods for PH, like persistence images, require refinement for complex biological data.

Purpose of the Study:

• To develop novel visualizations for persistent homology (PH).
• To optimize vectorization techniques for PH, specifically persistence images, by exploring different weighting strategies.
• To enhance the analysis and biological interpretation of complex spatial data from multiplexed imaging.

Main Methods:

• Proposed a novel visualization technique for persistent homology.
• Investigated and fine-tuned vectorization methods for PH, focusing on persistence images and varying weighting schemes.
• Applied the developed methods to spatial biological datasets.

Main Results:

• The novel PH visualizations provide new biological insights.
• Optimized persistence image vectorizations improve the analysis of complex spatial data.
• Demonstrated the utility of the methods on COVID-19 lung and lupus spleen datasets.

Conclusions:

• The proposed PH visualization and refined vectorization methods offer promising tools for analyzing multiplexed imaging data.
• These advancements facilitate deeper biological interpretation of complex spatial relationships between multiple cell types.
• The methods show potential for improving the analysis of diseases like COVID-19 and lupus.