CytoSpatio: Learning cell type spatial relationships using multirange, multitype point process models

Area of Science:

• Computational Biology
• Spatial Biology
• Bioinformatics

Background:

• Multiplexed fluorescence imaging advances enable detailed analysis of cell-cell interactions.
• Understanding spatial relationships is crucial for deciphering tissue function and disease.
• Existing methods often lack the ability to model complex, multi-type, multi-distance interactions.

Purpose of the Study:

• Introduce CytoSpatio, open-source software for modeling cell type interactions.
• Develop generative, multirange, and multitype point process models.
• Enable spatially realistic simulations of cellular relationships.

Main Methods:

• Constructing generative point process models to capture cell type interactions.
• Analyzing spatial relationships of five cell types across five tissue types.
• Validating models against a published dataset.

Main Results:

• Identified consistent spatial relationships within tissue types.
• Observed consistent clustering of proliferating T cells across tissues.
• Found tissue-specific attraction-repulsion dynamics between cell types like B cells and CD4-positive T cells.
• Generated synthetic tissue patterns from learned models.

Conclusions:

• CytoSpatio effectively models complex cell-cell spatial interactions.
• The software reveals conserved and context-dependent cellular relationships.
• CytoSpatio's generative capabilities offer new avenues for simulating tissue biochemistry and disease processes.