January 1, 2013 (Vol. 33, No. 1)

Michael D. Stadnisky, Ph.D
John Quinn, Ph.D.

Clinical trials and high-throughput flow cytometry produce large amounts of data, leading to an analysis bottleneck unaddressed by commercially available cytometry analysis software. As the throughput, applications, and content of flow cytometry have increased exponentially over the past several years, cytometrists are beginning to embrace cloud-based data-management solutions. The newest version of Tree Star’s FlowJo visualization and analysis software for flow cytometry advances research in flow cytometry by providing robust tools for analysis automation and a bridge to repository/cloud-based data.

Case Study 1: Automating Analysis of Clinical Trial Data

We recently analyzed a large dataset from a clinical trial of anti-IgE (omalizumab), which examined whether combination therapy with omalizumab and allergen immunotherapy for ragweed reduced seasonal allergic rhinitis compared with immunotherapy alone. The four-arm, randomized, double-blind, placebo-controlled study had nine time points, eight panels examining immune cell activation and one negative control panel—producing 10,991 flow cytometry standard (FCS) data files.

Manual organization, analysis, and report generation with these FCS files would produce a large bottleneck, slowing the progress of trial reporting and publication. However, with three key tools described in this article—group organization, templates, and command line—FlowJo provided an automation solution, addressing the bottleneck with a scalable solution that can:

  • Organize samples and batch analyze
  • Maintain consistent analysis for repeated SOP
  • Reduce downtime caused by manual analysis and report generation
  • Increase lab capability for high-throughput and high-content assays

Groups Organize Samples and Enable Batch Analysis

An intuitive user interface facilitates rapid setup of an analysis strategy, including statistics, hierarchical cellular subsetting through “gating” (boundaries which isolate events), and batch application to related samples using a patented groups structure. Data organization and management is a critical component in any data-analysis strategy, and groups organize related samples in a database-like structure, add metadata, apply analysis, and provide the underlying organization for batched reports.

Templates Execute Consistent Analysis for Repeated SOP

FlowJo has the ability to generate workspace templates, which saves sample organization, analysis strategy, and report output options without the flow cytometry data (FCS) files. Thus, when data is placed into a template, samples are organized into groups, analyses executed, reports generated, and analysis workspaces created and saved for review.

Command Line Reduces Time from Collection to Reports

The final piece of the automation puzzle comes through the ability to launch FlowJo without a user interface or human intervention. With a few lines of text, command line on Mac, PC, or LINUX is used to point FlowJo to the files to be analyzed, perform auto-compensation (signal processing), calculate gates and statistics, and produce output files such as batch spreadsheet and graph reports (Figure 1).

Figure 1. Automating flow cytometry analysis with FlowJo templates and command line: FCS files generated by a cytometer are analyzed using a FlowJo template through command line: Reports and analysis workspaces are saved to a user-set location for review.

Case Study 2: HT Analysis Combining Manual Gating and Clustering

Automated analysis of cytometry data has been advanced significantly by a variety of flow cytometry-specific clustering algorithms, which can be integrated as part of a workflow leveraging FlowJo’s gating, templates, and visualizations.

In this case, FlowJo is combined with two freely available clustering tools to perform a high-throughput analysis of T-cell intracellular cytokine assays after simian immunodeficiency virus (SIV) infection in order to:

  • Process many datasets in the same manner
  • Apply manual expert gates in a traditional manner
  • Apply automated gating at another level of the analysis to mine the data for additional information
  • Visualize cluster results

Data Preprocessing

By creating a preprocessing template, some simple and very common actions to the data can be applied. In this case T cells of interest were examined by setting up a hierarchy of gates (in order) on singlet cells, a simple clean-up gate to remove the cells on the axis, and a “magnetic” CD3+ gate.

This gating analysis was applied to all samples stained with the T-cell cytokine reagent panel (Figure 2), generating reports for QA that flagged outliers, and gated T-cell data (or further subsetting to analyze CD4+ and CD8+ T cells), and exported this pre-processed data for clustering analysis. Leveraging this template and command line, preprocessing analysis can be rapidly and consistently applied to all samples in this study.

Clustering Using GenePattern and R

GenePattern and the statistical program R offer freely available algorithms that perform clustering of flow cytometry data. In this case, we used two nonparametric clustering methods, SAMspectralclusterFCS in GenePattern and flowMeans in R, which both export .csv spreadsheet files representing the clustering results. This data can be drag-and-dropped or imported with a command line argument into FlowJo for association with the matching sample.

Currently, we are working to integrate these methods into FlowJo to enable those unfamiliar with the R environment to leverage the power of combining traditional gating analysis with automated clustering algorithms in a user-friendly environment.

Visualizing Clustering Results

Now, the clustering results can be analyzed and visualized like any other FlowJo-gated populations. When the cluster number file is dragged into FlowJo, a parameter for cluster number is created, and gates are made on each cluster number that separate the cells assigned to that cluster by SAMspectral (or any other algorithm) from all other cells.

Figure 2. Gating strategy and clustering algorithms: Gating on singlet CD3+ T cells using a hierarchal gating strategy in FlowJo. This data is exported for clustering in GenePattern and R, and clustering results brought back into FlowJo for analysis and visualization.

We can create a multigraph overlay of clusters meeting our threshold criteria (e.g., clusters that have more than 1% of the cells) as shown in Figure 3, allowing us to rapidly visualize cytokine-producing T cells.

Comparing Clustering and Manual Analysis

Finally, a new platform within FlowJo, FlowDx, can be used compare manual analysis (from experts and technicians) and clustering results from different algorithms. Using three statistical measures (match ratio, F-measure, and Davies-Bouldin), FlowDx evaluates consistency, accuracy, expert gating, and consensus analysis numerically to determine which classification methods agree and which do not. Critically, this allows for the assessment and comparison of manual and automated gating methods in high-throughput analysis.

Figure 3. FlowJo multigraph overlay of clustering results.


The throughput and content of flow cytometry assays is a data-management inflection point driven by the exponential increase in the number of fluorophores conjugated to antibodies used for cytometric interrogation of biological samples, an increase in the number of parameters that cytometers can detect, and an explosive growth in applications.

Cytometry is in a nascent stage of cloud/LIMS adoption; FlowJo Version 10 was built from the ground-up for cloud integration and provides a bridge to analyzing cloud-hosted data and automating analyses.

Critically, data analysis, report generation, and the entire FlowJo workflow are no different from analyzing files and saving reports locally. Thus, FlowJo can integrate with a data-management solution and provide seamless integration from the current paradigm of local file storage to accessing cloud/LIMS-based flow data for analysis and report generation.

Michael D. Stadnisky, Ph.D. ([email protected]), is application scientist coordinator and director of business development and John Quinn, Ph.D., is an application scientist at Tree Star

Previous articleBiggest Biotech Trends of 2012
Next articleLook, No Label! Label-Free MST