flowMerge

flowMerge optimizes preprocessing and data transformations for automated gating in flow cytometry to improve accuracy and consistency of cell population identification using the flowClust framework.

Key Features:

• Data Transformation Optimization: Evaluates generalized hyperbolic arcsine, biexponential, linlog, and generalized Box-Cox transformations within the BioConductor flowCore framework to enhance visualization and gating across fluorescence intensities and population variances.
• Parameter Optimization: Uses maximum likelihood criteria to optimize transformation parameters and reduce variability and misclassification (mis-gating) of cell populations.
• Performance Comparison: Systematically compares parameter-optimized transformations to default-parameter settings using real and simulated datasets and measures variations in cell population locations discovered through automated gating.
• Transformation Recommendations: Identifies parameter-optimized biexponential or generalized Box-Cox transformations as preferable for fluorescence channels and optimized generalized hyperbolic arcsine as advantageous for scatter channels in high-throughput contexts.
• Data-Specific Adaptability: Maintains a data-specific approach for selecting transformations to accommodate diverse datasets and experimental conditions.

Scientific Applications:

• Automated Gating Accuracy: Improves accuracy and consistency of automated gating workflows using flowClust in high-throughput flow cytometry.
• Immunophenotyping: Enhances identification of immune cell populations through optimized fluorescence channel transformations.
• Cancer Research: Supports more reliable cell population delineation in cancer research datasets.
• Stem Cell Analysis: Improves preprocessing for stem cell studies requiring precise population identification.

Methodology:

Implements a range of flow cytometry data transformations (generalized hyperbolic arcsine, biexponential, linlog, generalized Box-Cox) within the BioConductor flowCore framework; defines parameter-optimization criteria based on modeling assumptions about transformed data and employs maximum likelihood to estimate transformation parameters; evaluates the impact of transformations on automated gating performance with flowClust using real and simulated datasets and performs systematic comparisons between optimized and default parameter settings by measuring shifts in discovered cell population locations.