Cytometry and Antibody Technology

Cleaning up your flow data with algorithms – Bad data part 5

by | May 10, 2023 | Avoid Bad Data Series, Learn: Data Analysis, Software | 2 comments

It’s time to get back to the Bad Data series. I’ll use a data set I had to troubleshoot for one of my users to illustrate the importance of cleaning your data. We mentioned FlowAI and FlowClean in the first installment of this series, and this post will hopefully showcase how useful they can be.

My user presented me with the dot plot below where can see a decent amount of very negative events displayed on the PacBlue axis.


Pseudocolor plot of EYFP and CD4 PacBlue, where the events have been gated on FSCxSSC only.

As we’ve in the series, negative events may be cause by a compensation error. It should be easy enough to find an unsymmetrical display between some combination of fluorophores on an NxN plot. In this case however, no such problem is to be found.

NxN plot reveals no obvious compensation issue with the data.

This is where looking at your Time parameter is useful. This parameter is saved automatically during the acquisition and represents the number of events acquired per second on the instrument. In our case, the Time parameter reveals a number of issues with the way the sample was acquired – a number of clogs, inconsistent number of events acquired, and these very negative PacBlue spikes. With other data set, using color dot plots may be very useful to identify problematic data.

Pseudocolor plot of Time x PacBlue. Weirdness includes inconsistent number of events acquired over time, and downward spikes in fluorescence on the PacBlue parameter.

Here, I used FlowAI to clean up the data. There are several other algorithms that will perform a similar task, and I can’t claim to be able to explain the specifics of how each of them perform. But roughly speaking, they will find the data that doesn’t look right and remove it from the analysis. Once the data is cleaned up, the final analysis will look much better.

FlowAI removed the portion of the data that didn’t look like the rest


All is good in the world

It will be important to understand what the algorithm is doing to your data in order to figure out to what degree you can trust in the remaining data. Publications on your favorite tools are available on PubMed and elsewhere.

One way to mitigate your need for these solutions is to work on your sample preparation – optimize your single cell suspension, filter your samples, etc. If you need tips and tricks, check out our Flow Basics 2.0 class.

For this presentation, I ran FlowAI in FCS Express. See below for a very quick overview of how it was done. You will find a similar toolbox on other analysis platforms available at the CAT Facility such as FlowJo and OMIQ.

Avoid Bad Data Series

Read the full series of blog posts.

Flow Basics 2.0

A comprehensive online course that covers the protocol and optimization of staining cells, panel design, choosing controls, and instrument setup.


  1. You help me out with Flow Cytometry Data Analysis.
    Thank you

    • Send us your data by email and we’ll have a look. If we think it will require a sensible amount of our time, we’ll provide you with a quote to perform the work.


Submit a Comment

Your email address will not be published. Required fields are marked *

Search Blog