FCS Processing?

[jimbomahoney]

Hi all,

During training by Fluidigm, I understood that it was necessary to post-process the FCS files it generates.

However, this comment:

"If you have "compatible with FlowJo" checked (default when acquiring samples, wherein you don't use the dialog)" - from http://cytoforum.stanford.edu/viewtopic.php?p=3985#p3985

As well as the following from the Fluidigm training material:

"Software (6.0 versions and above) generates a two-data-segment FCS file:
• Randomized and presented: When the FCS reader opens the FCS file, the randomized segment of data is displayed by default.
• Non-randomized and hidden: The non-randomized data segment can be obtained by converting the FCS file into a TXT file."

Makes me think that the post-processing is perhaps only necessary if one wants to do different things (e.g. concatenation, clustering or altering the defaults).

I understood that post-processing was necessary to both normalise and randomise the data.

Is it the case whereby all FCS files must be post processed before analysis and that therefore I should do this for users before handing them the data, or that they need to install CyTOF software and do it themselves before analysis?

[mleipold]

Hi James,

I'm not exactly sure what you mean by "post-process the FCS file it generates".

If you have the defaults set up like Zach describes, then your raw, unchanged output FCS file *will* be Randomized, and will have the Per-parameter (rather than Global) scaling.


This Fluidigm "Randomized and presented" vs "Non-randomized and hidden" two-part FCS file is at least part of the reason why the Raw FCS files from the Helios are so big. After you normalize them (particularly with the Finck-style normalizers), the file size goes down significantly because the extra Fluidigm bits are chopped out.


Regarding normalization: if you are going to be doing any clustering, or Median/MFI-based analysis (fold-change for phosphos, etc), then you really do need to normalize. If you're just looking at Freq Parent, in my experience you often don't need to normalize, as you can just nudge your gates appropriately.

Therefore, if we're just doing a quick report to a customer for Freq Parent, we often don't normalize. This is because we've tried to stay fairly agnostic about the norm methods (Fluidigm vs Finck), for our customers. Some are hardcore Fluidigm (especially if they're comparing the new data against data acquired at another institution), some are hardcore Finck (especially if our or their Helios are more sensitive than the ones used by Fluidigm to create the Passport standard), and we leave that decision up to the customer unless they specifically request one method.

Since they can download Finck normalizers and Fluidigm normalizers, this has not be a barrier for them.


Mike

[jimbomahoney]

Hi James,

I'm not exactly sure what you mean by "post-process the FCS file it generates".

I mean run them through the CyTOF software "Process" function.

However, I think your answer covers this - i.e. yes, it's (usually) necessary to normalise and this must be done post-acquisition using either the CyTOF software or the Finck normaliser (the latest recommended version seems to be here - https://github.com/ParkerICI/premessa).

Thanks!