SwiftReg cluster registration automatically reduces flow cytometry data variability including batch effects

Area of Science:

• Computational Biology
• Bioinformatics
• Immunology

Background:

• High-dimensional flow cytometry data analysis is challenged by technical variations from cytometers, reagents, and operators.
• These variations obscure true biological differences, hindering accurate interpretation of experimental results.
• Existing correction strategies are often specific to variation types and cannot be easily automated.

Purpose of the Study:

• To introduce swiftReg, an automated computational method for reducing undesired variability in high-dimensional flow cytometry datasets.
• To specifically address and correct for batch effects, a common source of technical variation.
• To enhance the detection of genuine biological signals masked by technical noise.

Main Methods:

• Utilizes the SWIFT algorithm to generate a high-resolution cluster map of multidimensional flow cytometry data.
• Measures shifts in cluster positions between samples to derive registration vectors.
• Aligns subpopulations by adjusting cell parameter values based on registration vectors and applies batch control samples for correction.

Main Results:

• swiftReg effectively reduces undesired sources of variability, particularly batch effects, between samples.
• The method successfully aligns subpopulations across different samples and batches.
• Enhanced detection of biological differences is achieved by selectively reducing technical variation.

Conclusions:

• swiftReg provides an automated and effective solution for correcting technical variations in flow cytometry data.
• The method facilitates more accurate identification and analysis of biological differences in large datasets.
• Outputting registered data in standard .FCS files ensures compatibility with existing bioinformatics tools.