FAQ  •  Register  •  Login

Normalization without all bead channels?

Forum rules
Please be as geeky as possible. Reference, reference, reference.
Also, please note that this is a mixed bag of math-gurus and mathematically challenged, so choose your words wisely :-)
<<

ssivajothi

Contributor

Posts: 31

Joined: Mon Aug 01, 2016 2:31 pm

Post Fri Mar 09, 2018 6:20 pm

Normalization without all bead channels?

Hi all,

I am wondering (hoping) if it is possible to normalize CyTOF data with one of the bead analytes missing. I acquired a bunch of samples without one of the bead channels selected (stupid, I know) and the Fluidigm software won't let me normalize without that channel. Is there ANY way I can achieve? Please let me know. I am thankful for all advice!

Santhosh
<<

mleipold

Guru

Posts: 5792

Joined: Fri Nov 01, 2013 5:30 pm

Location: Stanford HIMC, CA, USA

Post Fri Mar 09, 2018 6:26 pm

Re: Normalization without all bead channels?

Hi Santhosh,

Unfortunately, none of the current methods (Fluidigm, MATLAB/Premessa) will work without all the five major EQ bead channels. And there's no way to "recover" data that was never collected in the first place (since it wasn't in your template).

The best chance you would have would be to talk to someone who could modify the normalization script(s) to leave out the missing channel.


Mike
<<

dtelad11

Master

Posts: 129

Joined: Mon Oct 31, 2016 6:26 pm

Post Fri Mar 09, 2018 6:28 pm

Re: Normalization without all bead channels?

I think you might be able to tweak the normalization scripts to ignore one of the channels.

Also, push comes to shove, just analyze your data without normalizing. How many samples are these? How many events?
<<

ssivajothi

Contributor

Posts: 31

Joined: Mon Aug 01, 2016 2:31 pm

Post Fri Mar 09, 2018 6:34 pm

Re: Normalization without all bead channels?

Hi

@Mike- that is what I assumed. I will try to contact the authors of the MATLAB normalizer to see if this is possible and/or if they would be willing to do this. Do you know of anyone else who would be able to accomplish this task?

@dtelad11- These were 4 runs with 12 barcoded samples each. So 48 samples in total with roughly 150,000 recovered events per sample. Is normalization not essential to interpret CyTOF results? I assumed findings would not be accepted from raw data.

Thank you,
Santhosh
<<

mleipold

Guru

Posts: 5792

Joined: Fri Nov 01, 2013 5:30 pm

Location: Stanford HIMC, CA, USA

Post Fri Mar 09, 2018 6:46 pm

Re: Normalization without all bead channels?

Hi Santhosh,

"Is normalization not essential to interpret CyTOF results?"

Short answer: it depends. I say this, based on splitting a sample and running the aliquots 8 hours apart (beginning and end of a long run day; yes, I can share the data if people are interested).


Longer answer: If you are simply comparing Frequencies of Parent or Cell Count, and your panel is reasonably robust (ie, decent separation between all populations), then, formally, no, normalization is not *strictly* required. In other words, if you can resolve your cell populations, those numbers should remain consistent.

However, if you're looking at Median Intensity Values (fold-change in phosphoflow, any kind of algorithm that uses intensity values to define clusters, etc), then it's best that you *do* normalize. This is because the decrease in signal intensity that you get over long runtimes may be enough to affect the numbers: either making extra clusters, or broadening your peaks enough to affect Median values.


Additionally: if you are running long barcoded samples, then you really *do* need to normalize somehow for good analysis. This is part of why the normalization beads were created: you can see in the Finck et al paper that when they run one sample for 2hr, they lose ~30% of their signal start to finish. Assuming the sample is evenly suspended/distributed, the end of the sample should look like the start of the sample, not 30% lower in signal.


Mike
<<

GregBehbehani

Master

Posts: 85

Joined: Tue Apr 12, 2016 10:17 pm

Location: The Ohio State University, Columbus, Ohio

Post Fri Mar 09, 2018 7:04 pm

Re: Normalization without all bead channels?

Hi Santhosh,

Ouch, that's painful. (But it could have been worse, if you ever barcode an experiment and then forget to record the barcode channels you'll know real pain.)

I would think you could make Rachel Finck's debarcoder do this, but you might have to add in an "empty" channel (corresponding to the missing bead channel) into the FCS file (you should be able to find someone who can do this for you in R), then when you manually select the beads just draw the boxes accordingly. You can then manually select the cut-off between bead events and cell events and I bet you'll be able to identify them fairly well.

Also, I would first just manually gate the beads and see if there's much signal drop. Often the actual signal change is small enough that the results will be the same with or without normalization.

Good luck,

Greg
<<

xiaoxiaoUC

Contributor

Posts: 20

Joined: Sun Feb 04, 2018 4:48 pm

Post Fri Mar 09, 2018 7:48 pm

Re: Normalization without all bead channels?

Mike,

Do you mind to share the data (you split one sample to 8 aliquotes and ran them in one day) with me?

Thank you so much!

Xiao
<<

ssivajothi

Contributor

Posts: 31

Joined: Mon Aug 01, 2016 2:31 pm

Post Fri Mar 09, 2018 8:09 pm

Re: Normalization without all bead channels?

Hi all,

@Mike- thanks for the detailed info. My run was long so I do expect some signal dropoff.

@Greg- I will look into a customized script and also just compare the bead signals. I think this much pain is enough for me, I don't even want to dream of missing those Pd channels!!

Best,
Santhosh
<<

mleipold

Guru

Posts: 5792

Joined: Fri Nov 01, 2013 5:30 pm

Location: Stanford HIMC, CA, USA

Post Fri Mar 09, 2018 10:43 pm

Re: Normalization without all bead channels?

Hi Xiao (and anyone else who might want it):

The data I mentioned earlier (sample stained, then split in half, one aliquot run first sample of the day, the second aliquot run last sample of the day) can be found at this Google Drive folder: https://drive.google.com/drive/folders/ ... sp=sharing

Let me know if it doesn't work. It contains the two times I did this, on a CyTOFv1 instrument back in 2014.

Both Raw and Normalized FCS files are included, as well as some summary data (Excel, Prism, PDF, etc).


Mike
<<

xiaoxiaoUC

Contributor

Posts: 20

Joined: Sun Feb 04, 2018 4:48 pm

Post Sat Mar 10, 2018 4:34 am

Re: Normalization without all bead channels?

Mike,

You're awesome! Your exps are very meaningful. Based on your data, can I understand that normalization by beads can deal with the signal intensity decrease issue by long runtime, although there still may be some batch effect that cannot be avoided?

I have another question (I'm sorry to be annoying): Did you do the metal spill over testing experiment before? What's your experience on that? Fluidigm people asked me to do it. But I think this should be done by them, not by us "customers".

Thank you! Nice weekend.

Xiao
Next

Return to CyTOF data analysis

Who is online

Users browsing this forum: No registered users and 12 guests

cron