Problems with expression level plots in Cytofkit

[AndrewD]

Hi,

We've generated data for the expression of 32 surface markers in PBMC. Ive ran the data through the cytofkit automated pipeline and its produced some fantastic results. However, we are having trouble visualising the expression of some of the markers on the tSNE plot. For example, CD3 and CD4 work, but then CD8 and CD19 do not (all dots remain blue as though all events have low expression); even though there is a shift of over 3 logs for the both the CD8 and CD19 signals by histogram. Also the cytofkit generated heatmaps show differing expression for these markers between the clusters.

Ive repeated the automated pipeline twice with the same results. Anybody else had these problems? What could be causing this?

Thanks for your help.

Andrew

[mleipold]

Hi Andrew,

Could you give us example data (CD3, CD4, CD8, CD19) in your tSNE plots and in flow bivariates (or histograms)? That might give us a better understanding of what you're describing.


Mike

[AndrewD]

Hi mike, thanks for the reply.

Ill try and explain a little clearer. Once the cytofkit automated pipeline has finished we load the .data file into the Cytofkit Shiny APP. We click on the tabs "Marker Panel" -> "Expression level plot" and we are presented with a tSNE plot (tSNE_1 vs tSNE_2 bivariate plot) with events coloured by expression of the selected marker. If we select CD3 we get a nice enrichment of expression in certain populations as expected but if we click on certain other markers such as CD19 we do not see a difference in colour (all appear to be low expressing). However if we then click on "Marker panel" -> "Expression level heatmap" tabs in the Shiny APP we can see that certain clusters are clearly enriched for CD19 expression.

What could be causing the problems with the Expression level plots and has anybody else had this?

Thanks for your help.

[MCOlivier]

Hi Andrew,

I think I already had that kind of "visual bias" in data analysis. If I correctly interpret what you see it can be related to either:
- the scale of the dual tSNE plot representation of your marker expression level is too tiny between max and min, making the differences between your events not visualizable (like when pre-gating on the CD45+ events, and seeing only "blue" events in Shinny APP, but with a scale from 4500 to 6000, meaning all those blue points are +)
- a very small Rphenograph cluster that is really more positive for your marker of interest when looking at the heat map, but that is spreaded in the tSNE map, so not visualizable instinctively.
- something else we can not deduce without any illustration of the problem :-p

Hope this helps ^^

Olivier

[AndrewD]

Ok, unfortuantely Ive no background in R so wouldnt know where to start in rescaling the coloured expression using this.

Ive imported the "analyzedFCS" file from cytofkit into cytobank and plotted tSNE_1_linear vs tSNE_2_linear and coloured for other parameters. Looks nice!!

Thanks

[MCOlivier]

Good idea to import those files in CytoBank for convenient rescaling/color choice.
However, in shinny APP, you can basically display the "all markers - scaled" to level/homogenize the color scales between markers.

[AndrewD]

The All markers(scaled) options does correct some of the problems and I can now see the expression in certain clusters on the tSNE plot now, but a problem still exists in that they are all blue (low expression) on the bluered scale.

The cytobank seemed to work much better.

[ErinSimonds]

@Andrew: A few versions ago on Cytofkit, there was a bug that affected plot color scaling. You can read about it in gory detail here. I'm using version 1.11.3 and the plots look good, although I admit I haven't done a direct comparison to Cytobank's plots of the same data.

I wonder if you might be using an old version. Can you run the command

  Code:

packageVersion("cytofkit")

in R and tell us the version number you're using?

Thanks,

Erin