Using CyTOF Data with FlowJo Version 10.0.7 Revised 2/3/14
Table of Contents 1. Background 2. Scaling and Display Preferences 2.1 Cytometer Based Preferences 2.2 Useful Display Preferences 3. Scale and transform modification 3.1 Arcsinh Transform 3.2 T-button 4. Tools in FlowJo for High-Dimensional Analysis 4.1 Drag and drop clustering 4.2 Multi-graph overlay 4.3 SPADE, Flow Means, and Other Data Mining Tools 4. Heat Mapping and Other Conditional Displays 5. Summary 1
1. Background The CyTOF is a mass cytometer produced by DVS Sciences (www.dvssciences.com). The name is an abbreviation for cytometry Time Of Flight, and the machine is used to conduct cytometry experiments on a much larger scale than can be achieved with a traditional fluorescence based cytometer. In a mass cytometry experiment instead of tagging antibodies with fluorescent molecules, the tags are isotopes of any element not found naturally in the biological system under study. Cells labeled with these antibody-isotope conjugates are then vaporized, ionizing the elements. This is done within a magnetic field so that the particles move across the magnetic field to a detector at a rate proportional to the size of the particle; the lighter the particle, the faster it moves. The measured time of flight is the time it takes particles to fly to the detector. What makes this process so useful is that there is very little noise associated with the ion tags. Each element can be resolved eliminating the need for compensation, and any element absent from a sample can theoretically be used as a reagent. FlowJo (www.flowjo.com) is the industry leading flow cytometric data analysis software. Its strength is in analyzing whole experiments encompassing many related samples. FlowJo has sophisticated tools that allow for generation of statistics, graphs, tables, web pages, mathematical models, and movies. FlowJo supports cloud deployment for excellent scaling of performance in a demanding, high-throughput environment, and can be integrated into Laboratory Information Management Systems (LIMS). 2. Scaling and Display Preferences FlowJo X inherently works with CyTOF data. There is nothing special that CyTOF user must do to get data on scale and all of the defaults scales are easily adjustable. Additionally we have included tools for high-dimensional data analysis. 2.1 Cytometer Based Preferences Preferences are the default settings for how FlowJo functions. There is a preference for most actions in FlowJo that can be controlled through the preference menu. That menu is opened by clicking on the heart in the top right hand corner of the workspace window, as shown in Figure 1. The heart opens to display a graphic tab for each part of FlowJo. One of the items that can be set is how FlowJo displays data from each type of cytometer. Click on the Cytometers tab, as shown in Figure 2, to view these settings. Figure 1 2
FlowJo adds a cytometer to the preference list the first time you add an FCS file with a unique $CYT keyword. Figure 3 shows the Cytometers preference list for a FlowJo user who has used data from many different cytometers. In the figure the CYTOF has been clicked on from the list on the left. The $CYT display shows that these settings will apply to all FCS files that have the value DVS- SCIENCES-CYTOF for the $CYT keyword. The default values are displayed at right. To change the default ranges for parameters displayed on either a log or linear scale, uncheck the box in front of the default scales. The numbers will un-gray, and the user may type any value they choose in. To set a particular parameter to be Always log, or Always linear, simply type the name of the parameter into the appropriate box. This is only necessary for parameters that are typically stored in the incorrect format on the cytometer. Figure 2 The Transformation Settings box allows the user to decide whether a transform should be applied to the data by default or not. If the Enable box is check for CyTOF data, an Arcsinh transformation (discussed more in section 2.1) will be applied by default. Figure 3 3
2.2 Useful Display Preferences Another set of preferences that may be of use to CyTOF users are the graph display preferences. From the Preference menu, these can be accessed by clicking on the Graphs icon, as shown in Figure 4. Figure 4 Figure 5 shows some of the items that may be of interest. Clicking on the menu indicated by bullet 1 allows the user to choose a particular type of graph to use as default. Some CyTOF users prefer to work with contour plots as an example. Another alternative is to smooth a pseudo-color plot, which can be achieved by clicking on the Smooth box just below this menu. Bullet 2 points at the checkbox that will allow outliers to be displayed if a graph that doesn t plot individual events, such as a contour plot or smoothed pseudo-color, is selected. Bullet 3 points to the location that allows the user to choose to display the target name only or the target and the metal probe. Bullet 4 allows the user to turn off the warning of Too many events on the axis for users who prefer to view untransformed plots. 4
Figure 5 3. Scale and transform modification Not all data will look best with the default settings. A user may want to change one parameter individually, or one sample individually. To change either the transform used, or the scale of the data in such a case the T button on any axis in a FlowJo graph window can be used. 3.1 Arcsinh Transform The default transform that FlowJo uses for CyTOF data is the Arcsinh. Unless the preferences are changed by unchecking the enable transform button, all CyTOF data receives this transformation for display. The Arcsinh transformation is the inverse of the sinh function of a data point. A plot of the Arcsinh function for a range of data is shown in Figure 6. The important aspects of this function are: It is linear in nature around zero, and allows for low end data to be visualized. Beyond a control point, the function becomes logarithmic in nature, allowing for increasing compression of positive populations. It is a mathematically defined function, and so it can be represented exactly as displayed in FlowJo in any other software, such as R. The bi-exponential transform, the default in FlowJo for most other cytometers data, is created using a lookup table, so the exact final equation to represent the transform is an approximation. 5
3.2 The T-button The T-button is on every axis in FlowJo. It can be used to change transforms, change ranges, or zoom in or out on data. In Figure 7 the T-button has been clicked to enable the user to switch from an arcsinh transformation to log, linear, or to customize the axis settings. Figure 6 Figure 8 displays the window that appears if you choose to Customize Axis. The list on the right shows which parameters will be affected by the change. The user can highlight multiple parameters if desired. The graphic is displaying the parameter for the axis from which the T-button was selected. The + and buttons on either edge of the axis expand or contract the displayed range. Below that (arrow 1) are a set of menus that allow you another place to change the transform type, or manually type in a range. Arrow 2 points at the box that allows for the width basis to be changed. This is not yet implemented for the Arcsinh function, but is for the bi-exponential, which produces a scale that looks similar to the Arcsinh. Arrow 3 points to a slider than can be moved to add extra negative decades. Figure 7 Increasing width basis pushes more of the low end data into the linear region, which occupies a fixed portion of the physical display space. The effect is that the low end data appears more compressed. Increasing the extra negative decades simply displays more space below zero. With CyTOF data this would just add extra white space to the plot. Figure 8 6
Figure 9 shows one CyTOF generated data file using different displays, transforms, and transform settings. Figure 9 For comparison, three plots from the CyTOF acquisition software are shown in Figure 10. The arcsinh transform is used for all of them with an increasing scaling factor. As the scaling factor is increased, the low intensity end of the data is increasingly compressed. A scaler value of 5 is the default and this is what FlowJo replicates. In FlowJo, switching to the biexponential transform replicates higher scaler values. Figure 10 7
4. Tools in FlowJo for High-Dimensional Analysis As well as displaying the data properly, FlowJo includes some analysis tools geared toward big data. For information on how any of the traditional FlowJo analysis tools work please see the FlowJo documentation at http://docs.flowjo.com/ or try the FlowJo tutorials at http://www.flowjo.com/home/tutorials/. 4.1 Drag and drop clustering Drag and drop clustering is the most general tool presented here. A user can apply a cluster analysis to FCS data using any platform and bring the results into FlowJo simply by dragging and dropping the results onto the FCS file in FlowJo. The tool must be able to produce outputs that are vectors of cluster numbers, so CSV or CLR file types are the most useful output. Figures 10 and 11 show an example of using this functionality. Sam Spectral, a spectral clustering algorithm available freely from the Gene Pattern Server (http://genepatternbeta.broadinstitute.org/gp/pages/login.jsf) flow cytometry analysis tools suite, was used on file A1. The output was saved as a CSV file. That file can be dragged and dropped directly onto the file within FlowJo, and gates representing each cluster will be created, as shown in Figure 11. Figure 10 Figure 11 4.2 Multi-graph overlay This tool can be used to either visualize multi-parametric data quickly, or used to visualize cluster results on multi-parametric data. Within the FlowJo layout editor, a user can drag and drop any FCS file in, or any sub-population identified using gating, right-click on it and immediately create plots of every two-parameter combination for the given population. Options also include creating histograms for every parameter, or 2D plots using the current Y-axis in every plot. Figure 12 shows how to access the NxN plots. Figure 13 shows a resulting plot. Figure 12 8
4.3 SPADE, Flow Means, and Other Data Mining Tools Any tool can be used to process data, and then displayed in FlowJo, but Treestar is also beginning to incorporate popular tools directly into FlowJo. The first two tools to be incorporated are SPADE and FlowMeans, two clustering algorithms. To access these tools, click on the Customize Ribbon tool, shown in Figure 14, which display all of the bands. Click on the Power band, shown in Figure 15, then drag it and drop it onto any Ribbon. Figure 16 shows the Power band dropped to the Configuration ribbon. Figure 13 Figure 14 Click on any population, open the Extra menu, and you can choose either Spade or FlowMeans from the list. Figure 15 9
Figure 16 Both of these tools require R to be present on the user s computer. How to install R, the necessary components for R, and how to use FlowJo-SPADE are covered in a separate document, available at http://docs.flowjo.com/vx/ advanced-features/spade-flowjo/. 4.4 Heat Mapping and Other Conditional Displays In the FlowJo table editor, statistics on any population can be conditionally formatted to highlight outliers, responders, etc. Three examples are shown in Figure 16; column 1 has been formatted with a heat map, column 2 has been formatted to highlight elements outside of 1, 2, and 3 standard deviations of the mean, and column 3 has been formatted to highlight files that produce statistics outside of a specified range blue for below the range and gold for above the range. These formats, data-linked plots, and other tools can be used to quickly identify nuances in large or complex data. Figure 16 10
5. Summary This whitepaper has provided information on how to scale, visualize, and apply multi-dimensional analysis tools to CyTOF produced data in FlowJo. The intent was to demonstrate that very little special consideration is needed to begin the analysis of CyTOF data and that the extent of the analysis time can be devoted to considering interesting tools to apply to mine the most information from this interesting data type. For FlowJo questions, please contact Treestar at flowjo@treestar.com. For CyTOF questions please contact DVS sciences at support.northamerica@dvssciences.com, support.europe@dvssciences.com, or (rest of world) support.row@dvssciences.com. FlowJo is Copyright Trustees of Leland Stanford Jr. University, 1996-1997 FlowJo is Copyright Tree Star, Inc. 1997-2014 CyTOF is Copyright 2013-2014 DVS Sciences Inc. 11