CytoGLMM: conditional differential analysis for flow and mass cytometry experiments.

Area of Science:

• Immunology
• Computational Biology
• Statistical Modeling

Background:

• Flow and mass cytometry are crucial for single-cell protein expression analysis in immunology.
• Current tools often compare many cell types, limiting focus.
• This study concentrates on analyzing a single cell type for more precise statistical modeling.

Purpose of the Study:

• To develop and evaluate statistical methods for differential protein expression analysis in flow and mass cytometry data.
• To address challenges posed by marker correlations and inter-subject heterogeneity in human immunology studies.
• To provide a robust workflow for analyzing single-cell data from cytometry experiments.

Main Methods:

• Employed two multiple regression strategies: a bootstrapped generalized linear model and a generalized linear mixed model.
• Assessed the robustness of these models against marker correlations and heterogeneity using simulated datasets.
• Illustrated the CytoGLMM R package and workflow for both paired and unpaired experimental designs.

Main Results:

• Both regression strategies maintained the target false discovery rate under medium marker correlations for paired experiments.
• Generalized linear mixed models demonstrated greater statistical power when the model was correctly specified.
• Detecting differences in unpaired experiments necessitates significantly larger patient sample sizes.

Conclusions:

• The developed approach effectively mitigates biases introduced by marker correlations in cytometry data.
• The methods provide safeguards against false discoveries stemming from patient heterogeneity.
• This enhances the reliability of differential protein expression findings in single-cell immunology research.