Predicting Cell Populations in Single Cell Mass Cytometry Data.

Area of Science:

• Immunology
• Computational Biology
• Data Science

Background:

• Mass cytometry by time-of-flight (CyTOF) enables high-dimensional single-cell analysis, crucial for identifying distinct cell populations.
• Current cell identification methods often rely on laborious manual annotation, hindering reproducibility and scalability, particularly in comparative studies.
• Existing automated methods may be complex, require prior biological knowledge, or are limited to canonical cell types.

Purpose of the Study:

• To introduce and evaluate a Linear Discriminant Analysis (LDA) classifier for automated cell population identification in CyTOF data.
• To demonstrate LDA's advantages in interpretability, reproducibility, and scalability compared to existing methods.
• To enable the identification of both known and novel cell populations for robust analysis of large-scale CyTOF datasets.

Main Methods:

• Application of a Linear Discriminant Analysis (LDA) classifier for supervised cell population identification.
• Benchmarking LDA against two state-of-the-art algorithms on four independent CyTOF datasets.
• Incorporation of a rejection option based on posterior probabilities to identify novel cell populations.

Main Results:

• LDA significantly outperformed two contemporary algorithms across four benchmark datasets.
• The LDA classifier demonstrated high performance in identifying both abundant and rare cell populations, providing accurate frequency estimates.
• LDA successfully identified previously unknown cell populations when a rejection option was integrated, showcasing its potential for discovery.

Conclusions:

• LDA provides a reproducible, scalable, and interpretable solution for automated cell population identification in CyTOF data.
• This method facilitates the analysis of large cohort studies by enabling accurate and consistent cell population profiling.
• LDA's ability to identify novel cell populations opens new avenues for explorative research in immunology and beyond.