How To Use A Threshold To Reduce Background Noise In Flow Cytometry
On most flow cytometers, the photomultiplier tube (PMT) is the interface between the fluidics system and the electronics system. It is the PMT that converts the photons emitted from the fluorochromes into the electronic current that is digitized and ultimately converted to the value stored in the listmode file.
Any stray photon of light or random electron emission from a dynode will cause a cascade, and ultimately a photocurrent. This is often known as dark current. The figure below shows the idealized idea behind this concept.
Figure 1: Stylized signal coming off a PMT showing the dark current and actual signals of cells passing the laser intercept.
As this figure shows, if each of these peaks is counted, there would be over 50 ‘events’ seen by the flow cytometer. Most of these events would be considered junk or debris.
Imagine if each of these events was recorded in the listmode file — how large would the file be?
To reduce this background noise in the system, we can use something called a threshold.
The threshold is a value that the signal must be above before the system will call a pulse an ‘event’.
If one enforces a threshold, the resulting pulse would look like this:
Figure 2: Stylized signal with a threshold added.
This threshold now reduces the signal from over 50 events down to just two. A much more manageable file size and analysis will be able to be performed by removing this noise.
1. Thresholding reduces background noise in the flow cytometer.
Thresholding is a powerful tool for reducing the signal caused by debris and dark current present in the flow cytometer.
Thresholding is a useful tool in reducing the debris that can overwhelm your datafile.
When computer storage was more expensive, and computers less powerful, it was much more heavily used.
Judicious use of threshold is warranted. However, it is also important to remember that if the flow cytometer doesn’t see an event, it doesn’t mean that the event is not present in the sample. This is especially true when sorting cells.
Here is an example of the effects of increasing the threshold on populations. These are CS&T beads run on a FACSAria, with an increasing threshold from 5,000 to 50,000 on the forward scatter parameter. A total of 20,000 events was collected for each file. Two gates (small and sort) are indicated.
Figure 3: Effects of increasing threshold on CS&T beads.
As can be seen by this data, increasing the threshold, decreasing the amount of debris seen in these beads, and the percentage of events in each gate, changes. These values are shown below.
Table 1: Percentage of cells in the two gates from Figure 3.
2. Thresholding removes smaller events.
Thresholding increases the percentage of target events in the datafile by removing the smaller events.
If one was performing immunophenotyping analysis, for example, the increased threshold resulting in a loss of the events in the ‘small gate’ would probably not be of concern. However, if one was to sort based on these different thresholds, a very different picture emerges.
The sort logic for this experiment is shown below, as generated in DIVA.
Figure 4: Sorting Strategy for Threshold Sorting Data.
After sorting the beads at different threshold levels, a post-sort analysis was performed on the instrument. Before the beads were placed back on the system, the threshold was reset to 5,000 and a wash was performed for two minutes, and the data from that wash shows the amount of background noise in the system.
A total of 89 events were observed in two minutes, with two of these events in the ‘small gate’ and none in the ‘sort gate’.
Figure 5: Results of a two-minute water wash showing the background noise in the system after sorting beads.
The results of the pre- and post-sorts are shown in figure 6. This is data from the ‘bright cells’ sort is shown, and the beads were sorted on purity mode.
Figure 6: Results of pre- and post-sort analysis at either 10,000 or 50,000 threshold on forward scatter. Post sort analysis was performed with a threshold of 5,000.
Table 2: Data from sort gates in Figure 6
The numbers tell the tale.
At a low threshold, the purity of the post-sort population of interest is extremely high, with only a minor contamination of the events in the ‘small gate’. However, when the instrument is blinded to the small events (via a high threshold), the post-sort analysis shows that there is significant contamination and much lower purity.
Since the system could not see the small events, it was not possible for these events to be excluded, thus the small events were sorted AT RANDOM into the collection tube, because the system did not abort those droplets where a small event was in the leading or lagging droplet.
While increasing the threshold will speed up the rate of acquisition of the events of interest, the effect of increasing the threshold must be weighed against the sensitivity of the downstream application.
If the cells are to be cultured, this debris may be tolerated. If the downstream analysis is a very sensitive technique, such as RNAseq, this debris might not be tolerated. It pays to be careful with the threshold to avoid surprises — like having your highly purified cell population contaminated with a host of unexpected genes (say 𝛃-globin).
Adding a threshold when acquiring flow cytometry data is like putting on sunglasses on a sunny day. It reduces the number of events by setting a bar that a signal pulse must clear before it is counted as an event. Depending on the importance of the data, the downstream applications for the data (or sorted cells) will dictate how critical the threshold is. Threshold wisely and practice proper sample preparation to reduce the debris in the tube to ensure the best outcome.
To learn more about how to use a threshold to reduce background noise in Flow Cytometry, and to get access to all of our advanced materials including 20 training videos, presentations, workbooks, and private group membership, get on the Flow Cytometry Mastery Class wait list.