FLAME: Analyze your own data  Print-icon

Use the FLAME modules to define and characterize discrete populations in flow cytometric data.

Preliminaries

You will need flow cytometric data in a .zip file. The flow files must be in .txt or .fcs format; all files must be in the same format.

Example data file: SMALL_phospho.lymphgated.fcs.zip.

Steps

  1. Login to the public GenePattern server: http://genepattern.broadinstitute.org/gp/.
  2. If you do not have a GenePattern account, you can register on the login page.

    Result: GenePattern displays its main page.

  3. Run FLAMEPreprocess:
    • Open FLAMEPreprocess. Click this link to display the module page in GenePattern. The parameter fields are empty and ready to populated with your data.
    • Open FLAMEPreprocess with example data. Click this link to display the module page in GenePattern and have GenePattern automatically populate the parameter fields with the example data.
    What it does: Tips:

    Performs a series of preprocessing operations on flow cytometric data files, including column/channel selection, bi-exponential transformation, and optional live-cell gating.

    • The input file is a .zip file containing flow files in .txt or .fcs format. The files should be all .txt or all .fcs, but not both.
    • FLAMEPreprocess documentation

  4. Run FLAMEMixtureModel:
    • Open FLAMEMixtureModel. Click this link to display the module page in GenePattern. The parameter fields are empty and ready to populated with your data.
    • Open FLAMEMixtureModel with example data. Click this link to display the module page in GenePattern and have GenePattern automatically populate the parameter fields with the example data.
    What it does: Tips:

    Clusters each preprocessed sample data file over a range of possible cluster numbers.

    • The input file is the output file from FLAMEPreprocess: a .zip file containing preprocessed flow sample files in .txt format.
    • Depending on data size and parameters, this module may take hours to run. The skew distributions require more computation time. If the populations are expected to be roughly symmetric, use t or normal distributions for faster response. If more than ten samples are to be analyzed, analyze a few samples to determine the optimal density distribution and number of clusters. Then, analyze all samples using that density distribution and a small range of cluster numbers.
    • FLAMEMixtureModel documentation

  5. Run FLAMEChooseOptimalClusterNumber:
  6. What it does: Tips:

    Determines the optimal number of clusters for each sample based on the range of cluster numbers provided to FLAMEMixtureModel.

  7. FLAMEMetacluster:
    • Open FLAMEMetacluster. Click this link to display the module page in GenePattern. The parameter fields are empty and ready to populated with your data.
    • Open FLAMEMetacluster with example data. Click this link to display the module page in GenePattern and have GenePattern automatically populate the parameter fields with the example data.
    What it does: Tips:

    Takes the data samples, which have been optimally clustered into subpopulations, and matches the subpopulations so that a given population can be identified uniformly across all samples.

    • The input files are: (1) the output file from FLAMEChooseOptimalClusterNumber, a .zip file containing the optimal mixture modeling result of each sample, and (2) the sample class names file, a two-column text file where the 1st column contains sample names (without .fcs or .txt filetype appended) and the 2nd column contains the corresponding class names.
    • When samples in a dataset belong to different classes, the module performs "metaclustering" in two steps: within-class metaclustering and cross-class metaclustering. When all samples belong to a single class, the module performs only within-class metaclustering.
    • FLAMEMetacluster documentation

  8. Download and unzip the FLAMEMetacluster result file.
  9. FLAMEMetacluster generates a .zip file that includes the final clustering results for each sample at its optimal cluster number, where each population/cluster is represented using a consistent color and position across all samples. Download and unzip this generated result file.

    In the next step, you will view the clusters for a given sample. To do so, you need the *.membership.txt and *.parameters.txt file for that sample. These files are in the FLAMEMetacluster .zip file, which you have now downloaded and unzipped.

  10. Run FLAMEContourViewer.Pipeline to view clusters for a given sample:
  11. What it does: Tips:

    Runs the FLAMEContourDataGenerator module to generate the data required to display a 3-D contour plot of the clusters in a given sample. Then, opens the FLAMEViewer to display a 3-D scatterplot and 3-D contour plot of the clusters in the sample.

    • The input files are the *.membership.txt and *.parameters.txt files for the sample of interest. Both of these files are in the .zip file generated by FLAMEMetacluster, which was downloaded and unzipped in the previous step.
    • FLAMEViewer documentation provides instructions for using the viewer.

GenePattern tips

Input and output files

The FLAME modules are designed to be used as a pipeline, where the output file from one module is used as the input file to the next module. GenePattern makes this easy:

  1. Run the first module. GenePattern displays the module and its output files in the Recent Jobs list on the home page.
  2. Click the icon next to the output file that you want use. GenePattern lists the modules that will accept this output file as an input file.
  3. Select the next FLAME module. GenePattern displays the module page and automatically populates the input file parameter field with the selected output file.

Creating a pipeline

After running the FLAME modules to analyze your data, you might want to create a pipeline that captures the modules, parameters, and data files that you used. GenePattern makes this easy as well:

  1. Analyze your data by running FLAMEPreprocess, FLAMEMixtureModel, FLAMEChooseOptimalClusterNumber, and FLAMEMetacluster. GenePattern lists the modules and their output files in the Recent Jobs list on the home page.
  2. Click the icon next to the .zip file generated by FLAMEMetacluster, which contains the final results. GenePattern displays an option menu.
  3. Select the Create Pipeline option. GenePattern creates the pipeline required to reproduce this analysis results file.
Updated on May 20, 2011 20:00