Is Warning $PnR NA is invalid while running flowCore::read.FCS safe to ignore?

[denvercal1234GitHub]

Hi there,

I encountered the Warning of $PnR NA is invalid when running flowCore::read.FCS on my spectral flow FCS files.

Note that I first exported the FCS files as Channel values, then use Spectre::write.files() to convert them into FCS files (ImmuneDynamics/Spectre#149), so that I can use flowCore::read.FCS() because ultimately I want to use the peacoQC package which only accepts FCS files.

The warning was discussed at #192 by @cciccole and @gfinak, but I am unclear what the final advice was. Would you mind helping me address this warning?

Thank you.

Code:

ff <- flowCore::read.FCS("/Users/clusteredatom/.../Analysis_LiveSinglets/Transformed_ChannelValues/channel_to_FCS_forpeaCoQC/channelFCS_LiveSinglets_Each_AfterMerge/channelFCS_FileNo_1.fcs", truncate_max_range = FALSE, transformation = FALSE)

Ouputs:

Warning: $PnR NA is invalid. If it is a numeric string, this could be because it is larger than R's numeric limit: 1.79769313486232e+308. The assigned $PnR value will be imputed from the maximum value of the data in this channel.
> 
> ff
flowFrame object 'channelFCS_FileNo_1.fcs'
with 98073 cells and 63 observables:
                       name                   desc     range  minRange  maxRange
$P1                   FSC-A column FSC-A from da..       639         0       638
$P2                   SSC-A column SSC-A from da..       639         0       638
$P3                  TOX1_2 column TOX1_2 from d..       766         0       765
$P4                   FOXP3 column FOXP3 from da..      1003         0      1002
$P5                PERFORIN column PERFORIN from..       801         0       800
...                     ...                    ...       ...       ...       ...
$P59              Comp-PE-A column Comp-PE-A fro..       NaN         0       NaN
$P60                 FileID column FileID from d..         2         0         1
$P61             FileName.y column FileName.y fr..         2         0         1
$P62 Time..SSC.H.subset.L.. column Time..SSC.H.s..     98074         0     98073
$P63 Time..SSC.H.subset.L.. column Time..SSC.H.s..        99         0        98
331 keywords are stored in the 'description' slot
> 

[mikejiang]

This doesn't impact the data that has been parsed from your fcs file per se, these instrument range/maxRanges are imputed by the maximum data value as warning message suggested.
Inaccurate instrument range may or may not affect your workflow down the road depending on what tools you will be using, how they utilize this range information, for example ggcyto uses instrument range to set plot range under certain circumstance.

but I do see you still have an invalid range value ($P59) in your flowframe object, which is concerning

you can print the original keywords to check what has been exported by your Spectre::write.files() function
lapply(1:dim(fr)[2],function(i)keyword(ff)[[paste0("$P",i,"R")]])

also print the data range for $P59 to see what's going on
range(ff, "data")

[denvercal1234GitHub]

Thank you very much, @mikejiang!

I think the P59 having NaN here is because we ran the experiment with 100 wells. Each well has the same backbone antibodies, but for PE, each well has a different PE-conjugated antibody. So, when I exported the data from the Aurora spectral cytometer, for some wells I assigned a name for the PE channel, but for other wells, I left blank.

So, actually, for this particular FCS file, the PE channel was not P59-Comp-PE, but CD1-PE which does have values and not just NaN.

I will try to go back to the naming and keep the PE channel just blank for all files, and see if this warning still appeared and will let you know.

[ghar1821]

Realised I replied to the wrong github thread. See here: ImmuneDynamics/Spectre#149

[denvercal1234GitHub]

Thank you so much for your time @ghar1821!

I will try `divide.by="FileName" and let you know, and check on QUESTION 3 and let you know.

And, yes please, having a guide on how to incorporate the following packages to our Spectre package would be very useful and increase the packages' usage a lot:

• infinityFlow package (https://github.com/ebecht/infinityFlow ; ebecht) upstream of Spectre.
• peacoQC (https://github.com/saeyslab/PeacoQC , AnneliesEmmaneel)
• flowVS to decide optimal cofactors per channel - which at the moment, here is how I do it, incorporating the steps by Modified workflow to use flowVS-transformed FCS in Spectre clustering workflow? HdBraanker/Spectral_Flow_Workflow#3 by HdBraanker
• Your tutorial (Interoperability: How to convert Spectre data.table to object compatible with CATALYST package? ImmuneDynamics/Spectre#92) on how to convert data.table to SCE object would also be useful for people wanting to use some features of the catalyst package (https://github.com/HelenaLC/CATALYST , HelenaLC).

[ghar1821]

@denvercal1234GitHub I'm copy pasting your reply above to the Spectre github issue thread. Save cluttering this issue as I originally replied to your question in the wrong thread!

[denvercal1234GitHub]

This was the issue.

Solution = change the marker names before exporting the data from cytometer into FlowJo and subsequent R packages.