Generating Quantitative Cell Identity Labels with Marker Enrichment Modeling (MEM).

Area of Science:

• Computational Biology
• Immunology
• Oncology

Background:

• Multiplexed single-cell techniques like mass cytometry generate high-dimensional data for cell population characterization.
• Traditional analysis methods often require expert knowledge or can yield unexpected groupings with novel algorithms.

Purpose of the Study:

• To introduce Marker Enrichment Modeling (MEM) as a method for quantitative cell population labeling.
• To demonstrate MEM's utility with both expert and machine learning-derived data.

Main Methods:

• MEM quantifies feature enrichment in cell populations relative to a reference.
• The MEM R package calculates enrichment values, generates labels (heatmap or text), and assesses label similarity.
• Analysis protocols are demonstrated using immunology and oncology datasets.

Main Results:

• MEM generates objective, quantitative labels for cell populations.
• The method effectively integrates with diverse analytical approaches, including expert and machine learning outputs.
• MEM facilitates the characterization of both known and novel cell populations.

Conclusions:

• MEM offers a robust framework for defining cell population identity and novelty.
• This approach enhances the interpretability of complex, high-dimensional single-cell data.
• MEM supports computational and expert-driven analyses in fields like immunology and oncology.