Stabilizing your unmixing: drift

In this set of posts, I'm going to look at how to go about stabilizing your unmixing, with the aim of making things a bit easier. In part, this is going to be a discussion of how often we need to be running controls and how to do it if you're not running controls every time. I'm going to start by saying that, philosophically, I'm always in favor of running controls every time since this provides you with more information to potentially get the best unmixing. That's not always practical, though.

What are some scenarios where you might not be running the controls with the samples? Let's say you're running your big flow panel that you've developed. You've run a bunch of samples on Monday, including the single-stained controls. The unmixing looks adequate. But, you don't manage to finish running everything and have to run the rest the next day or the day after--is that a problem? Alternatively, you might be dealing with clinical samples. You receive a large set early in the week and run those, but then you get an unexpected sample on Friday--do you prepare all 40+ controls to run that one sample, or do you just add it to your experiment because you're in a rush and didn't plan to do this anyway?

Today's post is meant to illustrate why I do not recommend running new samples using the same unmixing matrix across multiple days. What I mean by that is just adding new samples to an existing experiment, assuming the spectral signatures will be stable and that there's no change in how the cytometer detects these fluorophores. I'm going to show what I think is the best-case scenario for this, using an instrument that is working properly and QC'd, using the same instrument settings, using the same lot of antibody and focusing on stable fluorophores. This isn't about changes in the fluorophores, it's about changes in how the instrument detects the fluorophores. If you add in changes in fluorophores, you can expect much bigger problems.

Variability in how single-stained beads appear after unmixing using an unmixing matrix from Day 0 on the Cytek Aurora.

Here's an example of that drift occurring on the ID7000.

Variability in how single-color controls appear when unmixed using an unmixing matrix from the first day on the Sony ID7000

These single-color controls are all prepared identically, using the same lot of antibody. The machine has been QC'd, passed and is working within the expected parameters. The instrument settings are fixed and this is using Standardized Mode. Moreover, these are fairly or very stable fluorophores. The fact that this unmixing matrix contains several fluorophores clustered together in the ~500nm range does amplify the effect. Basically, we notice smaller changes in the unmixing because we're trying to prise apart signatures that are very similar. In a panel with well separated fluorophores or with minimal complexity, this will be less of an issue. Given that you're doing spectral flow and reading this blog, that's probably not what you're after.

The controls are the most important part of any experiment, flow cytometry or otherwise. Without good controls, we can't interpret the data.

Purple-crowned Fairy, Costa Rica