gaussNorm is a tool used for batch normalization. gaussNorm uses landmarks (peaks) on a per-channel basis that has been manually inputted by the user, to address batch effects. Batch effects are considered to be non-biological variation between groups of samples.
Hahne, F., Khodabakhshi, A., Bashashati, A., et al. Per-channel Basis Normalization Methods for Flow Cytometry Data. Cytometry, 77A:121-131 (2009). https://doi.org/10.1002/cyto.a.20823
1. Prepare your Data
1.1 Perform any Compensation/Unmixing and Scaling Tasks
Prepare your data as you usually would for flow cytometry analysis. Find links to useful resources below:
1.2 Perform Cleanup Gating and Subsampling Tasks
It is recommended to perform cleanup gating and subsampling to exclude debris, doublets, dead cells, etc from the normalization task. Debris, doublets, dead cells, etc can lead to a sub-optimal gaussNorm normalization.
To learn more about performing gating and subsampling, please our our How to Perform Gating and Subsampling resource page and our Performing Manual Gating article.
2. Add gaussNorm Task to your Workflow
Click Add new child task and select gaussNorm from the task selector. In this example, we have applied compensation to the files. We have subsampled to live cells for our gaussNorm task.
Your exact workflow branch may look different than the example above, depending on Steps (1.1) and (1.2). The important thing is that your workflow follows a logical ordering of tasks.
3. Setup and Run gaussNorm
3.1 Select Files and Features
Select the Files you want to correct for batch effects. If you have control files, you can include them in the selection. Select the Features you want to include. Note that the physical parameters are not included.
3.2 Enter gaussNorm Settings
Comma-separated landmarks per channel: Input the number of landmarks (peaks) in each channel separated by commas. The order of the peaks should match the list of features selected. For example, in the picture above the comma-separated list is 1,1,1,2 and so on; would mean the BUV395-A | IFNg has 1 peak, BUV661-A | CD127 has 1 peak, BUV737-A | IL2 has 1 peak, BUV805-A | CD3 has 2 peaks, and so on.
To easily determine the number of peaks your features have, you can create histograms of each channel in a Figure task.
3.3 Run
Click Run gaussNorm. This will take you to the Status tab and you can watch the progress. However, you are free to go back to your workflow or do whatever you please while this runs in the cloud. The status can also be seen in the Workflow itself, or you can have an email sent to you when it is completed.
4. Review your Results
Go to the Results tab to review your Results. Here, I have chosen to download the before_and_after_norm.pdf to view how gaussNorm has normalized the data.
5. Continue to Build your Workflow
If you are satisfied with the batch normalization results, then you can simply continue to build this workflow branch and the normalized Features will be available in downstream tasks.
Useful Tip: Change Channels to Reflect Normalized Values
You can change the feature channels to reflect the normalized values.
Go to the Results tab. In the Result Columns, open the Rename Result Columns.
Click Configure Names.
Select all Result Columns. Click Submit.
Type in the Primary Feature Name of the channel in the New Name Column. Click Submit.
To easily determine the Primary Feature Name of a channel, you can have a second OMIQ window open that has the Setup tab of the gaussNorm task open, as pictured below. To learn more about working with two windows, please see our support article Two Window Analysis in OMIQ.
In the image above, the window with the gaussNorm Setup tab open has the Select Features column visible. The syntax of feature names that have a primary and secondary name is primary name|secondary name. For example, the feature BUV395-A | IFNg, the primary name is BUV395-A and the secondary name is IFNg.
Select the Files you want the name change to be applied to. Click Submit.
The features downstream of the gaussNorm task will now only contain the normalized features. This means that your normalized features will replace your original features in downstream tasks. Your normalized parameters will take on all qualities of your original parameters, including any scaling choices.