flowVS: channel-specific variance stabilization in flow cytometry.

Area of Science:

• Biotechnology
• Bioinformatics
• Statistical Analysis

Background:

• Comparing heterogeneous cell populations across biological conditions is crucial in flow cytometry (FC).
• Standard statistical methods require stabilized within-population variances for accurate comparisons.
• Mean-variance correlation in FC data necessitates preprocessing for variance stabilization.

Purpose of the Study:

• To introduce flowVS, a novel algorithm for variance stabilization in flow cytometry data.
• To address the challenge of mean-variance correlation in fluorescence-based FC measurements.
• To provide a robust method for enhancing the reliability of cell population comparisons.

Main Methods:

• Implemented an inverse hyperbolic sine (asinh) transformation for each fluorescence channel.
• Optimally selected transformation parameters using Bartlett's likelihood-ratio test to achieve homogeneous variances.
• Applied the optimized transformation to all samples within a dataset for consistent variance stabilization.

Main Results:

• flowVS effectively removes mean-variance dependence from raw FC data.
• Demonstrated homogeneous within-population variances across fluorescence channels.
• Showcased superior variance stabilization compared to existing methods like flowTrans, flowScape, logicle, and FCSTrans.
• Validated flowVS performance on microarray data, achieving results comparable to state-of-the-art VSN software.

Conclusions:

• Variance homogeneity is essential for uniform feature extraction and comparison of cell populations in FC.
• flowVS provides an effective solution for variance stabilization in both FC and microarray data.
• The algorithm facilitates improved comparison and alignment of phenotypically identical cell populations across diverse samples.
• flowVS and associated datasets are publicly available via Bioconductor for broader scientific use.