Page 1 of 1

Difficulty with Nolan stand-alone debarcoder

PostPosted: Fri Jul 07, 2017 7:34 pm
by mleipold
Hi all,

I've been trying to use the Nolan stand-alone debarcoder on a Helios file (I'm not using the Fluidigm BC kit), and I haven't been able to get to the debarcoding part.

SingleCellDebarcoderR2013b_MacOSX, version 1.
https://github.com/nolanlab/single-cell ... -FCS-Files

I created my CSV key of Masses in columns, BC combinations in rows. I made sure to not leave the upper left cell blank. I can open the debarcoder, select my CSV key, and select my FCS file, but I then get nothing but a pop-up saying "Check your barcode key". It does not progress further.


1. Since raw Helios files output with *.FCS and some other software has had problems, I tried changing the filename to *.fcs. No effect.
2. I tried doing the "Paste special"/"Values" recommendation in case somehow Excel was messing that up. No effect.
3. The example CSV key has "102" as a BC channel, but just in case there was a parameter naming issue between CyTOF2 and Helios, I also tried "Pd102", since that's what FlowJo lists as the metal/mass. No effect, nor any of the other CSV variations that I tried.

Cytoforum is telling me that "The extension csv is not allowed." for attached files, so the relevant FCS, Xlsx, and CSV files can be found at: https://drive.google.com/drive/folders/ ... sp=sharing


I have to assume that it's something really trivial.......



Thanks,
Mike

Re: Difficulty with Nolan stand-alone debarcoder

PostPosted: Mon Jul 10, 2017 2:15 pm
by ErinSimonds
Hey Mike,

Usually the hard part for people is getting the Matlab runtime installed, so you're already past the worst of it. Try the sample barcode key and sample FCS file from the Github page, just to get a known starting point where everything works:

https://github.com/nolanlab/single-cell ... d_file.fcs
https://raw.githubusercontent.com/nolan ... de_key.csv

...then replace that sample file with a Helios file (encoded with the Fluidigm 20-plex kit), if you have one, so we can eliminate weird filename, folder name, and file extension issues.

Some ideas on why it might not like your barcode key:

1. You are using Y89 as a channel. It's also possible that it is looking for a 3-digit number (i.e. 089).

2. You only have two barcoded populations, and it is usually run on files with 10-20 populations. Try inventing 8 intermediate populations using your markers (for a total of 10).

ES

Re: Difficulty with Nolan stand-alone debarcoder

PostPosted: Mon Jul 10, 2017 6:35 pm
by mleipold
Hi Erin,

Thanks for your suggestions.

1. I was able to successfully debarcode the Github Sample file, using the Github Sample csv key. Minimally, I got output files that were read by FlowJo and seemed to make sense in the little gating I did.
2. I removed the 89Y from my CSV key (leaving one BC as 11000 and one BC as 00111). I was then able to get the debarcoder to read my Helios *.fcs file and execute to create output files B1 and B2. Again, they opened in FJ and made sense in a little checking of markers/gates.
3. I then replaced "89" with "089" (using Text as the text format type, otherwise leading "0" was removed). In this case, I still got the "Check your barcode key" and doesn't progress.

I have added those relevant XLSX and CSV files to the Drive folder with the previous files.

So, as far as I can tell, it appears to be an 89Y-related issue, where the Debarcoder doesn't want to read (or accept?) 89Y as a potential BC. Thanks for the suggestion to try that, as I never would have thought to do it.

BTW: I have not yet tried adding 194 or 198 (ie, something other than Pd) to see whether it will execute or not....


Mike

Re: Difficulty with Nolan stand-alone debarcoder

PostPosted: Mon Jul 10, 2017 10:52 pm
by mleipold
Hi all,

I happened to have a barcoded file from a 2-of-8 scheme from a couple years ago: 104, 106, 108, 110, 113, 115, 194, 198. I made the corresponding 28 BC key, knowing that a couple of the reagents (104, 106, 108) had failed and therefore there were several BC's which should have N=0 cell events.

I have uploaded the file in question, and the corresponding BC key (CSV) to the same Drive folder.

The MATLAB debarcoder was able to handle all of the BC combinations, even though not all of them were Pd102-110. Therefore, my previous trouble appears to be limited to the presence of 89(Y) in the original recent BC key.


Erin: do you know if the MATLAB debarcoder is still being maintained/updated? Granted, right now, there aren't many things <100 mass units that we would want to barcode with, but that may be different in the future.


Thanks for your help,
Mike

Re: Difficulty with Nolan stand-alone debarcoder

PostPosted: Wed Jul 12, 2017 8:40 pm
by ErinSimonds
Hey Mike,

Good work on the tests and finding a bug! The Matlab debarcoder is not under active development, so I'm not sure this will ever get fixed. Rachel Finck coded this before she left the Nolan Lab.

Federico Gherardini has been working on a replacement coded in R: https://github.com/ParkerICI/premessa

As the Github page states, "The software is still under heavy testing." However, it does work if you'd like to try it. Also, it's unlikely that the 89Y problem will carry over to the R version, because it's probably some weird hardcoded thing in Matlab.

Another option for you is to rename the channel in the FCS header from Y89 to Y189. Since actual 189 channel is probably empty in your pannel, I don't think this little hack would cause any issues. Then the Matlab debarcoder should handle it just fine. You can use CytofCore for this:
https://github.com/nolanlab/cytofCore

- Erin

Re: Difficulty with Nolan stand-alone debarcoder

PostPosted: Wed Jul 12, 2017 10:55 pm
by BjornZ
Thanks for troubleshooting!

I opened an issue and fix here:
https://github.com/nolanlab/single-cell ... der/pull/2

(Unfortunately I don't have a way to compile this for MacOS or Windows right now, so no help unless you have a full Matlab install. I'm seeing if anyone I know has Matlab on other platforms.)

Re: Difficulty with Nolan stand-alone debarcoder

PostPosted: Mon Jul 24, 2017 11:19 pm
by mleipold
Hi Erin, Zach (and Federico and Rachel, on Github):

As a follow-up: I can confirm that Premessa (the R debarcoder) does work with the file I had which possessed the 89Y BC channel.

However, it's definitely still under construction.....


Mike

Re: Difficulty with Nolan stand-alone debarcoder

PostPosted: Fri Sep 15, 2017 3:13 pm
by markrobinsonca
Hi all,

A bit late to the game here, but just to say that the R/Bioconductor CATALYST package also has a debarcoder:

https://bioconductor.org/packages/CATALYST

Best wishes, Mark

Re: Difficulty with Nolan stand-alone debarcoder

PostPosted: Fri Dec 15, 2017 10:12 pm
by ezunder
Also late to the game, but I wanted to add:

My lab is maintaining an updated version of the Finck/Zunder Matlab debarcoder:
https://github.com/zunderlab/single-cel ... Debarcoder
https://www.ncbi.nlm.nih.gov/pubmed/27897009

We got help from Jake Eccles, who updated the code to deal with 89Y barcode reagents.

Best,
Eli