Data scaling (also known as data transformation) is the process in which the original data value is transformed to a new value. This is necessary to perform on data whenever the analysis of the data is not appropriate in the original units of the data. This is the case for cytometry data, where positive signals can spread over massive numeric distances while dim or negative signals do not. Cytometry data are fundamentally log-like, and thus are re-scaled accordingly to be interpreted linearly.
To learn more about data scaling, please see our guide Understanding Data Scaling.
1. Open a Scaling Task
Click Open on the Scaling task.
Default workflows in OMIQ will have a scaling task added. Depending on whether the files have a spillover matrix stored in them, your default workflow might look different.
Adding a second scaling task downstream of an existing scaling task is not ideal. The second scaling task will be transforming already scaled data. This can result in events with very small values. If you need a second scaling task below the first scaling task to set Display Bounds, you can set the Scaling Type to None (linear).
2. Manually Set Scaling Cofactor
The image above shows a view of the scaling task. Each column is explained below. You also see in the table the currently displayed Y axis and X axis of the plot denoted by the eye icon Y or eye icon X respectively. On the right column, you can see a plot. You can change the X and Y axis by clicking on the arrows here. Note that if you click on a feature under the feature name column, this will be displayed as the X axis on the plot.
Feature Name: Column of available features in the dataset.
Scaling Type: Choose between None (linear), Arcsinh, or Log10.
Caution should be taken when choosing Log10 scaling type since this cannot handle negative numbers. Therefore, this might not be appropriate for your data and may introduce artifacts and cause errors in your analysis. We believe that Arcsinh is best for the field of cytometry as a whole. The Arcsinh function is ubiquitous and simple. You can read more about this in our article Understanding Data Scaling under the section There Are Many Scaling Functions, All of Them Similar.
Cofactor: Arcsinh can be considered a "hybrid scale" as it behaves linearly around 0 and logartihmically further away from 0. The cofactor defines how much of your data around 0 is displayed linearly. This is analogous to the "width basis" in biexponential scaling.
Setting the cofactor manually may take some trial and error. Setting the cofactor too high will result in over compressed events around 0 and may start to pull positive events towards 0. Setting the cofactor too low will result in under compressed events around 0 and may result in elongation of the negative events, appearing as a splitting artifact around 0. You can read more about this in our article Understanding Data Scaling under the section Modifying Scaling Settings and the Effect on Data and Choosing the Right Scaling Settings.
Min: This is the minimum display bound for the feature.
Max: This is the maximum display bound for the feature.
The display bounds does not effect the numeric properties of the data. This is the way to move the events on the plot to minimize unnecessary white space.
You can Bulk Edit the scaling type and cofactor. Select the Features for bulk edit. Select a Scaling Type and input a Cofactor as necessary. You can also bulk edit the Min and Max Display Bounds.
The scaling task does not remove events below or above the min or max display bounds. This is primarily for visualization and a way to move the events on the plot to minimize unnecessary white space.
3. Set the Scaling Cofactor Using a Spreadsheet
Select Copy Settings and click on System Clipboard. This copies the scaling settings to the system clipboard in a spreadsheet format.
Open a spreadsheet editor and paste the scaling settings.
The image above is a pasted scaling task in a spreadsheet editor. The columns represent the different columns found in the scaling task, including the Min Z and Max Z columns for the Z coloring axis. Edit the table as desired. Here, we have changed the scaling cofactor for BUV395-A, BV570-A, and APC-Fire 810-A
Copy the cells in the spreadsheet editor. Ensure that you copy the column headers as well.
In the Scaling task in OMIQ, click on Paste Settings and select System Clipboard.
Paste the Settings from the System Clipboard and click Submit.
A confirmation appears that all the scaling and display bounds are successfully pasted.
4. Automatically Set the Scaling Cofactor
Click on Autoscale.
Select a File to use as the basis for the autoscaling algorithm. Click Submit.
An ideal file to use would be the most representative staining of your complete panel. This template file is used to set every non-linear channel in the task. For this reason, we recommend using a fully stained file that most represents your files.
A confirmation appears that autoscaling is complete.
You can override the cofactors set by autoscale, if required, by manually entering them. See Step 2 for more.
Tip: You Can Set the Z Coloring Minimum and Maximum in the Scaling Task
Click on Toggle Z. This will toggle the Z min and Z max color range for each feature set.
The image above shows a scaling task with the Z coloring min and max displayed as Min Z and Max Z. The current Z coloring is denoted by the eye icon Z. You can select the feature for the Z axis in the setting Z axis. Note that in the Bulk Edit feature, the Display Bounds cannot be change when the Z axis is toggled.