Doublw width issue

[jobagins]

Hi all
We observe in few exeriments samples with two width populations - please check the picture. Digging in, those doesnt seem to be doublets and cellular phenotypes remain exactly the same for both of those. Fluidigm suggested to analyze as one population.
Interestingly, we did one experiment on Tuesday in which double width appears, then repeated yesterday on the same exact donor/protocol and we don't have double width populations anymore.
Any thoughts?

[mleipold]

Hi Joanna,

Could you show some Marker Intensity plots for the two Width populations? Like, Ir, CD3, CD14, CD20, etc?

Since Event Length changes in your first plot, I would find it surprising if Marker Intensity *didn't* change......


Mike

[jobagins]

Hi Mike

Thank you for your help. Please find here the screenshot, I am happy to share fcs. We observe this phenomenon on some occasions and often when re-run the same donor using the same Abs panel we don't observe it again
Please let me know if any suggestions
Thank you
Joanna

[mleipold]

Hi Joanna,

Just to confirm: in these plots, the Marker is on the Y-axis and the Event Length is on the X-axis? It's hard to tell without axis labels (and scales.....)

If so, you have a significant Freq difference between Top and Bottom. For example, CD3+ is 65% in Top, while it's 43% in Bottom. If Top and Bottom were truly the same, then they should have the same Freq.

Could you put the Event Length on a Linear scale (maybe 0-80)? The log (or asinh) scale makes it hard to see what's going on.


One possibility: if your nebulizer develops a partial clog to where it's deflecting the spray slightly, you can get your ion clouds to where they're not focused on the cone orifice (think throwing a dart at the edge of a dartboard, rather than the bullseye where you want it).

In most cases I've seen, this causes an overall drop in signal intensity across all channels (fewer ions getting in = fewer ions getting counted). However, you could imagine an edge case where the deflection is big enough to be noticeable in the data but not so bad as to give you *no* cells or *no* signal.

If that's the case, then the differences in Freq you see between Top and Bottom could be that not all of the cell events satisfy the EL and/or LCT lower thresholds and therefore aren't getting Counted (ie, cells aren't being written to the FCS file). If so, reprocessing the data with lower EL and/or LCT thresholds would get the Freq Parent to be consistent between Top and Bottom.

My *assumption* would be that the Bottom set is being undercounted. If so, the the Bottom would have a sharp edge on the left side of the EL distribution, whereas the Top would have a hotspot that would be a smooth distribution.


I also wonder if there might be *two* things happening here:
1. The Width issue, which in your first plots seems to be consistent over time (ie, Top and Bottom appear at all timepoints).
2. Whatever is causing the drop in EL vs time in your first plot, about 2/3 of the way through the Time. But which is happening to *both* Top and Bottom, consistently.


Aside: Brian and Adeeb's MMB book chapter ( https://dx.doi.org/10.1007/978-1-4939-9454-0_2 ) goes into the definitions of Residual, Offset, Width, and Center. I think there's a Fluidigm notice about it too, but I could find the MMB chapter first.

If I'm reading Fig 7 correctly, Width = 20 * s.d. of the ion peak width (as measured in Pushes). So, having populations with two different Widths would imply having populations with two different ion peak widths.


Mike

[jobagins]

Thank you so much Mike for looking into this, it's very helpful. Here the updated image.
Population ratios does change yet signal remains similar. Its impossible to notice those 2 populations if gaussian parameters are not used for cleanup.
Initially we suspected barcoding troubles, however we see this phenomenon in non-barcoded samples as well.

[mleipold]

Hi Joanna,

Thanks for the figures....that clarifies what we're looking at.

One last question: I see that you have at least 2 Ir populations. Do you see these two Ir populations (as clearly defined as here) when you *don't* have two Width populations?

If you gate on just one of the Ir populations, how does that affect Top and Bottom Width? In other words, is Top Width the upper Ir population, and Bottom With the lower Ir population?


I'm running out of ideas......

Mike

[jobagins]

Yes absolutely, I was also initially thinking if I gate on singlets on Ir I will get rid of it which is not the case. Please see.....

Yes we see these two Ir populations as clearly defined when when we don't have two width pops.

I will lean towards acquisition itself being a problem but would like to understand it.

Thank you so much for your input

[desireeBCRT]

Hi Joanna,
we have had similar issue recently. Similar to what you describe DNA-vs-DNA looked normal, but event length was effected. I only noticed that something was wrong because the events running in the rainplot looked strange. Some EQ beads were much "longer" and also the DNA staining of some of the cells was "longer". In the data file we had two populations of beads and also cells when looking at event length.

The problem was resolved when I changed the injector - it was a bit damaged around the edge and had white residue at the tip. I cannot say for sure that this was the reason, because I did not check again with this injector, but before and since then we did not see the issue again.

Best
Desiree

[jobagins]

Thanks Desiree!

We have a feeling that it still might be independent of injector and investigating it, if I have any interesting findings will share it

Thank you!

Joanna

[rickd598]

Hi

Did anyone managed to solve this issue, we have a similar problem with a double amplitude signal, with the weaker signal amplitude population correlating with poor barcoding. Staining of phenotyping markers looks similar in both populations with the exception of cd19 and cd20, where there appears to be a loss of these cells in the low amplitude population rather than a reduction of marker expression. We are currently assuming an issue with barcoding or fixation, but are unsure. Any input is appreciated