Statistical modeling and analysis of cell counts from multiplexed imaging data.

Area of Science:

• Computational biology
• Biostatistics
• Pathology

Background:

• Multiplexed imaging technologies allow detailed spatial mapping of cells in healthy and diseased tissues.
• Existing statistical models are insufficient for comparing tissue cellularity across different sample groups.
• Accurate statistical analysis is crucial for understanding tissue composition in health and disease.

Purpose of the Study:

• To develop and validate statistical models for analyzing cell count distributions in multiplexed imaging data.
• To identify statistical tests that enhance power for differential abundance testing.
• To address challenges in analyzing highly aggregated cellular data in tissue samples.

Main Methods:

• Development of two novel statistical models for cell count distributions.
• Application of models to imaging mass cytometry data from lymph node, COVID-19 lung, and Hashimoto disease tissues.
• Comparison of statistical power of new tests against traditional rank-based tests.

Main Results:

• The developed models accurately describe cell count distributions, linking parameters to field of view size and cellular properties like density and spatial aggregation.
• Identified statistical tests show improved power for differential abundance testing compared to rank-based methods.
• Spatial aggregation significantly impacts statistical power, necessitating larger sample sizes for highly aggregated cells.

Conclusions:

• The proposed statistical models provide a robust framework for analyzing multiplexed imaging data.
• A stratified sampling strategy is introduced to reduce sample size requirements when dealing with aggregated cells.
• These advancements facilitate more powerful and efficient comparisons of tissue composition across sample groups.