CoChat

Optimizing transformations for automated, high throughput analysis of flow cytometry data

Open access
Finak, Greg, Perez, Juan-Manuel, Weng, Andrew, Gottardo, Raphael · 2010 · BMC Bioinformatics

The DOI matches its Crossref record; 27 of the 27 references it cites were found in a registry or an index.

Abstract

BACKGROUND: In a high throughput setting, effective flow cytometry data analysis depends heavily on proper data preprocessing. While usual preprocessing steps of quality assessment, outlier removal, normalization, and gating have received considerable scrutiny from the community, the influence of data transformation on the output of high throughput analysis has been largely overlooked. Flow cytometry measurements can vary over several orders of magnitude, cell populations can have variances that depend on their mean fluorescence intensities, and may exhibit heavily-skewed distributions. Consequently, the choice of data transformation can influence the output of automated gating. An appropriate data transformation aids in data visualization and gating of cell populations across the range of data. Experience shows that the choice of transformation is data specific. Our goal here is to compare the performance of different transformations applied to flow cytometry data in the context of automated gating in a high throughput, fully automated setting. We examine the most common transformations used in flow cytometry, including the generalized hyperbolic arcsine, biexponential, linlog, and generalized Box-Cox, all within the BioConductor flowCore framework that is widely used in high throughput, automated flow cytometry data analysis. All of these transformations have adjustable parameters whose effects upon the data are non-intuitive for most users. By making some modelling assumptions about the transformed data, we develop maximum likelihood criteria to optimize parameter choice for these different transformations. RESULTS: We compare the performance of parameter-optimized and default-parameter (in flowCore) data transformations on real and simulated data by measuring the variation in the locations of cell populations across samples, discovered via automated gating in both the scatter and fluorescence channels. We find that parameter-optimized transformations improve visualization, reduce variability in the location of discovered cell populations across samples, and decrease the misclassification (mis-gating) of individual events when compared to default-parameter counterparts. CONCLUSIONS: Our results indicate that the preferred transformation for fluorescence channels is a parameter- optimized biexponential or generalized Box-Cox, in accordance with current best practices. Interestingly, for populations in the scatter channels, we find that the optimized hyperbolic arcsine may be a better choice in a high-throughput setting than current standard practice of no transformation. However, generally speaking, the choice of transformation remains data-dependent. We have implemented our algorithm in the BioConductor package, flowTrans, which is publicly available.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1186/1471-2105-11-546

Cites

27 of 27 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 Flowchart of the analysis pipeline. Flowchart describing our analysis pipeline. a) Procedure for analyzing FSC vs SSC channels. Standard data analysis procedures are depicted by path 1, whereas procedures applying parameter-optimized transformations are depicted by path 2. b) Procedure for analyzing fluorescence channels. Standard procedures are depicted by path 1). Procedures utilizing optimized transformations are depicted by path 2. The default transformation depicted in 1) is the generalized arcsinh with default parameters (a = 1, b = 1, c = 0), as defined in the flowCore package. Normalization follows transformation in b) to ensure that the transformed data are on a common scale.
Figure 1 Flowchart of the analysis pipeline. Flowchart describing our analysis pipeline. a) Procedure for analyzing FSC vs SSC channels. Standard data analysis procedures are depicted by path 1, whereas procedures applying parameter-optimized transformations are depicted by path 2. b) Procedure for analyzing fluorescence channels. Standard procedures are depicted by path 1). Procedures utilizing optimized transformations are depicted by path 2. The default transformation depicted in 1) is the generalized arcsinh with default parameters (a = 1, b = 1, c = 0), as defined in the flowCore package. Normalization follows transformation in b) to ensure that the transformed data are on a common scale.CC BY · PubMed Central
Figure 2 Simulation study results. Results of transformations on simulated data. a) A single simulated sample is shown as a series of bivariate dot plot projections. Data are presented on the original scale (top row), and on the scale of the inverse biexponential transform (bottom row). Points represent individual events. b) Boxplots representing the distribution of misclassification rates of flowMerge models with K = 9 components fitted to the simulated data set under different transformations. c) Intra-cluster variability measured as the total sum of squared deviations for metaclustered populations identified by flowMerge under different transformations. d) Example bivariate projections of metaclusters for untransformed data (top row), default biexponential (second row), optimized biexponential (third row), optimized generalized arcsinh (fourth row), and default generalized arcsinh (fifth row). Corresponding metaclusters were selected where possible. Metaclusters are labeled as +/+, -/-, -/+ for artificial markers A and B. Ellipses represent 90th quantile contours of subpopulations.
Figure 2 Simulation study results. Results of transformations on simulated data. a) A single simulated sample is shown as a series of bivariate dot plot projections. Data are presented on the original scale (top row), and on the scale of the inverse biexponential transform (bottom row). Points represent individual events. b) Boxplots representing the distribution of misclassification rates of flowMerge models with K = 9 components fitted to the simulated data set under different transformations. c) Intra-cluster variability measured as the total sum of squared deviations for metaclustered populations identified by flowMerge under different transformations. d) Example bivariate projections of metaclusters for untransformed data (top row), default biexponential (second row), optimized biexponential (third row), optimized generalized arcsinh (fourth row), and default generalized arcsinh (fifth row). Corresponding metaclusters were selected where possible. Metaclusters are labeled as +/+, -/-, -/+ for artificial markers A and B. Ellipses represent 90th quantile contours of subpopulations.CC BY · PubMed Central
Figure 3 Visualization of FCM data under different transformations. Visualization of flow cytometry data under different transformations. a) The FSC and SSC dimensions of a representative sample transformed using the biexponential, generalized arcsinh, generalized Box-Cox, and linlog transformations with optimized parameters. Untransformed and default-parameter generalized arcsinh and biexponential transformed data are shown for comparison. Some parameter-optimized trans-formations (biexponential and generalized Box-Cox in this example) improve visualization and resolution of the lymphocyte cell population when compared to default or untransformed data. b) Comparison of a fluorescence channel data under different parameter optimized and default transformations. In this example, the optimized biexponential and optimized linlog improve the resolution of the two populations in the CD19 vs CD5 channels, compared to the default generalized arcsinh or default biexponential. Points represent individual events, contours represent the two-dimensional kernel density estimate of the data.
Figure 3 Visualization of FCM data under different transformations. Visualization of flow cytometry data under different transformations. a) The FSC and SSC dimensions of a representative sample transformed using the biexponential, generalized arcsinh, generalized Box-Cox, and linlog transformations with optimized parameters. Untransformed and default-parameter generalized arcsinh and biexponential transformed data are shown for comparison. Some parameter-optimized trans-formations (biexponential and generalized Box-Cox in this example) improve visualization and resolution of the lymphocyte cell population when compared to default or untransformed data. b) Comparison of a fluorescence channel data under different parameter optimized and default transformations. In this example, the optimized biexponential and optimized linlog improve the resolution of the two populations in the CD19 vs CD5 channels, compared to the default generalized arcsinh or default biexponential. Points represent individual events, contours represent the two-dimensional kernel density estimate of the data.CC BY · PubMed Central
Figure 4 Variation in cell subpopulation locations. Variation in population location measured as the sum of squared distances between population centers for a) lymphocyte populations in the FSC vs SSC channels. b) metaclusters of cell populations in the fluorescence channels.
Figure 4 Variation in cell subpopulation locations. Variation in population location measured as the sum of squared distances between population centers for a) lymphocyte populations in the FSC vs SSC channels. b) metaclusters of cell populations in the fluorescence channels.CC BY · PubMed Central
Figure 5 Metaclusters in the scatter channels. Metaclusters of lymphocyte populations in the FSC and SSC dimensions for four optimized and three default transformations. We observe little difference between the transformations in the forward and side scatter dimensions for gating lymphocyte populations, suggesting that the primary benefit in the FSC vs SSC dimensions is for data visualization. Contours represent the 90th quantiles of the cell subpopulation distributions.
Figure 5 Metaclusters in the scatter channels. Metaclusters of lymphocyte populations in the FSC and SSC dimensions for four optimized and three default transformations. We observe little difference between the transformations in the forward and side scatter dimensions for gating lymphocyte populations, suggesting that the primary benefit in the FSC vs SSC dimensions is for data visualization. Contours represent the 90th quantiles of the cell subpopulation distributions.CC BY · PubMed Central
Figure 6 Metaclusters in the fluorescence channels. Metaclusters of cell populations defined in the fluorescence channels of the lymphoma data under different transformations. Only the primary cell populations are shown for comparison (CD19+/CD3-/CD5+ and CD19-/CD3+/CD5-). The number of metaclusters for the transformed data is shown in brackets. The number of cell populations in a metacluster is shown above the plot. Contours represent the 90th percentiles of the cell subpopulation distributions.
Figure 6 Metaclusters in the fluorescence channels. Metaclusters of cell populations defined in the fluorescence channels of the lymphoma data under different transformations. Only the primary cell populations are shown for comparison (CD19+/CD3-/CD5+ and CD19-/CD3+/CD5-). The number of metaclusters for the transformed data is shown in brackets. The number of cell populations in a metacluster is shown above the plot. Contours represent the 90th percentiles of the cell subpopulation distributions.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 27 references

Cited by

See all 99

Metadata

Authors
Finak, G., Perez, J.-M., Weng, A., & Gottardo, R.
Affiliations
  • Fred Hutch Cancer Center (US)
  • Montreal Clinical Research Institute (CA)
  • Terry Fox Research Institute (CA)
Corresponding author
Finak, Greg
Venue
BMC Bioinformatics
Published
2010-11-04
Volume
11
Issue
1
Pages
546-546
ISSN
1471-2105 (ISSN-L)
OpenAlex
W1976192231
Cited by
99
References
27
Citations by year
2026: 4 · 2025: 5 · 2024: 9 · 2023: 6 · 2022: 7 · 2021: 7 · 2020: 12 · 2019: 14 · 2018: 8 · 2017: 6 · 2016: 2 · 2015: 5 · 2014: 4 · 2013: 4 · 2012: 4
FWCI
1.93
Type
journal article
Language
en
Open access
Licence
  • cc-by
  • creativecommons.org/licenses/by/2.0 (tdm)
Locations
  • BMC Bioinformatics (publishedVersion, cc-by)
  • Europe PMC (PubMed Central) (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
Funders
  • National Institutes of Health
  • Canadian Institutes of Health Research
  • National Institute of Biomedical Imaging and Bioengineering (R01EB008400)
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Biophysics › Cell Image Analysis Techniques
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Single-cell and spatial transcriptomics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Gene expression and cancer classification
Keywords
automated gating, high-throughput flow cytometry, parameter optimization, maximum likelihood estimation, Bioconductor
MeSH terms
Algorithms, Cells, Data Interpretation, Statistical, Flow Cytometry, Databases, Factual

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Optimizing transformations for automated, high throughput analysis of flow cytometry data. https://app.cochat.ai/papers/10.1186/1471-2105-11-546
The paper
Finak, G., Perez, J.-M., Weng, A., & Gottardo, R. (2010). Optimizing transformations for automated, high throughput analysis of flow cytometry data. BMC Bioinformatics. https://doi.org/10.1186/1471-2105-11-546

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.