GammaGateR: semi-automated marker gating for single-cell multiplexed imaging

Area of Science:

• Computational Biology
• Biotechnology
• Bioinformatics

Background:

• Multiplexed immunofluorescence (mIF) enables multichannel protein imaging for cell-level spatial analysis in tissues.
• Current automated cell phenotyping methods for mIF data lack consistency and often require subjective manual evaluation.
• Manual thresholding for marker gating in mIF analyses is time-consuming and prone to variability.

Purpose of the Study:

• To develop an evaluable, semi-automated algorithm for consistent and reproducible marker gating in mIF data.
• To introduce GammaGateR, an R package designed for interactive marker gating on segmented cell-level mIF data.
• To provide a robust tool that overcomes the limitations of existing automated and manual mIF analysis methods.

Main Methods:

• Developed GammaGateR, an R package utilizing a closed-form gamma mixture model for marker gating.
• Implemented user-specified constraints for a consistent, slide-specific model fit in mIF analysis.
• Compared GammaGateR against unsupervised methods using colon and ovarian cancer datasets.

Main Results:

• GammaGateR demonstrated highly similar results to manual annotation benchmarks (silver standard).
• The package effectively identified biological signals, such as spatial interactions between CD68 and MUC5AC cells.
• Phenotype probabilities from GammaGateR accurately predicted survival in ovarian cancer patients when used with machine learning.

Conclusions:

• GammaGateR is an efficient R package that enhances the replicability of marker gating in mIF studies.
• The tool significantly reduces the time required for manual segmentation and analysis of mIF data.
• GammaGateR facilitates more reliable cell phenotyping and biological discovery from complex imaging data.