The Daily Dongle

For the Love of Transformation

From time to time the question comes up, "why does my data look slightly different in FlowJo than it does in my acquistion  software?". Well, sometimes vendors don't adhere to data standards, and sometimes the math that drives data display is slightly different. What follows should help clear up some hows and whys of data display, offer tips getting data looking the same and/or help clarify why we display data the way we do. We focus on Diva, and will be following up soon about Accuri, Coulter and other data files.

Why do my plots look different in FlowJo compared to BD FACSDiva?

This is due to the default transformation settings in the two software packages. Traditionally, without transformation, cells with negative fluorescence would be considered to have zero fluorescence and would begin to 'pile-up' on the axis.  Fluorescent baseline subtraction error during acquisition is a fundamental of flow cytometry and is one reason why 'negative' fluorescence is observed. Traditional logarithmic scaling compresses the channels of visual space as the scale increases. This leads to 'visual misrepresentations' of dispersed populations in the lower ranges of the logarithmic scale and condensed populations in the higher ranges of the scale.  By transforming the data, the scale is compressed in the lower range, typically from 1-10 or 1-100, leading to a more accurate visual representation of fluorescence units in the low range of the scale as compared to the higher range of the scale.  Hence, using transformation provides a more precise visualization tool in comparing populations with low  fluorescence versus those with high fluorescence.


BD FACSDiva Users:

1) Exporting from Diva should ALWAYS be done as FCS files and NOT as an experiment.  Exporting your data as an experiment will alter  your data and is designed for use in Diva for analysis.

2) FCS 3.0 data will be automatically transformed when loaded into FlowJo according to the default Preference settings. The automatic transformation can be turned OFF within the Preferences and also the transformation values can be changed here.

3) The Width Basis that Diva uses for its default transformation is set to -100.  FlowJo uses a default Width Basis of -10.  It is FlowJo's opinion that compressing 100 channels of space around zero may (in some cases) compact 'dim' populations into negative populations.  This compaction is less likely to occur when using a Width Basis of -10.  You can keep the Width Basis at -10 by default and increase it per parameter if needed. However, if you want your transformation to appear the same way in Diva and FlowJo, change the default transformation Width Basis in the FlowJo Preferences to -100.

4) If your data is exported from Diva as FCS 2.0, no transformation can be applied unless a compensation matrix is created in FlowJo.

5) Although FCS 3.0 data is automatically transformed using the default settings, transformation of any parameter can be changed using the 'T' button next to the parameter in the graph window (PC) or under Platforms>Biexponential Transformation (Mac).

Further FlowJo Biexponential Transformation information/examples can be found at:

Width-Basis Titration
Biexponential Cramps
Mac Version and Biex Transforms
Dongeloids Guide to Transforms

Quad Gates Using FJML

This is an interesting application of FJLM/Vertex gating in FlowJo V9.2 - driving the center position of a quad gate based on another gate's position or summary stat. 

Let's say you like using quad gates - and you want to tie the center position to more less where the negative cells end on x/y axes. Here's how I assembled that. 

FlowJo 7 and CSV export

This question just came up on PUCL, so I probably should mention the few ways of handling CSV data in FlowJo 7. This pretty much applies for the version 9 as well.

 

The simplest use case is you have a bunch of FCS files and want to convert all of them into CSV files:

- load files into FJ workspace
- select all the files (ctrl+a)
- click worskpace menu, Export/Concat
- switch type to CSV, and let it rip - this will generate a folderfull of CSV's at your specified destination.

 

Another common one is "I only want a specific gate exported". this is also batch-able in FJ - assuming your gate is group-owned by many samples, you can repeat the above steps with slight variation - instead of selecting all the root nodes (CTRL+A), click once on the gate you want to export, then click Edit -> Select Equivalent Nodes. Then click workspace->export/concat as above.

 

I should also mention the opposite direction, going from some kind of CSV file into FSC so it can be read into FlowJo is also possible. For this, we have a free stand-alone java (all platforms) conversion utility I wrote with Seth and Jay posted a few entries back .

Orderly Export of Tables from FlowJo PC to Excel

Export a table from FlowJo 7.x into Excel  with the following fun steps:

1. Drag populations and statistics into Table Editor.


2. Click the Batch button and open the Excel tab.
   
   
3. Click the drop-down menu and select Excel.      (CSV will work too)


4. Choose a file name and destination. Check Open file after creation.


5. If Excel is set to open .xls (and CSV)  files on your computer, the file will open in Excel.

You can also export to Excel after making the table inside FlowJo. (Batch button, then the FlowJo tab)
From the top of the FlowJo table, select  FILE > Save as... > Excel

- Don Gull

Roll Your Own FCS Files Part 2

Seth, Jay and I have worked on a little utility we use internally which may also be of interest to some of you, as an alternative to the mac-only FJTEXTDATA trick described here.

This tool converts text files (csv, etc) into FCS files. Our application of this tool is making synthetic data with known properties to test FJ's ability to calculate statistics. I also use the tool to parse various information from our web logs or email statistics. The general idea is to bridge "generic data" and flowjo's user interface for data mining and batching. 

 

Download TextToFCS (roughly 50MB, a MUCH smaller version is available if you don't mind compiling it in command line.)