Hi John,

It will absolutely work to collect your compensation controls and samples without compensation applied and then generate and apply a compensation matrix in FlowJo post-hoc. I do this routinely with a 10-color panel. Since your single-color compensation controls have to be acquired and analyzed in order to generate the compensation matrix, they are not themselves compensated at the time of acquisition (i.e., you can't apply a matrix that you haven't already generated!). The key thing here is that your compensation controls are single-color - so for example if you want to compensate FITC out of the PE channel, you need a "FITC only" control that the software "knows" only contains FITC. Thus, for that sample, if there is PE signal, then it must be spillover, and will be subtracted out by the software. Just run a single-color sample for every color in your panel and you'll be able to compensate your experimental samples in FlowJo. If you want to be sure its working, you could compare the manually-generated compensation to the one generated by FlowJo - it should give you the same answer. I also first learned the basics of flow with the FACSCalibur/CellQuest - manual compensation is pretty convenient when you only have four colors, but if you move to an instrument that allows bigger panels, manual comp becomes pretty cumbersome, and it ends up being easier and probably better to compensate using FlowJo.

Hope this helps, 

Max

Hi John,

It will absolutely work to collect your compensation controls and samples without compensation applied and then generate and apply a compensation matrix in FlowJo post-hoc. I do this routinely with a 10-color panel. Since your single-color compensation controls have to be acquired and analyzed in order to generate the compensation matrix, they are not themselves compensated at the time of acquisition (i.e., you can't apply a matrix that you haven't already generated!). The key thing here is that your compensation controls are single-color - so for example if you want to compensate FITC out of the PE channel, you need a "FITC only" control that the software "knows" only contains FITC. Thus, for that sample, if there is PE signal, then it must be spillover, and will be subtracted out by the software. Just run a single-color sample for every color in your panel and you'll be able to compensate your experimental samples in FlowJo. If you want to be sure its working, you could compare the manually-generated compensation to the one generated by FlowJo - it should give you the same answer. I also first learned the basics of flow with the FACSCalibur/CellQuest - manual compensation is pretty convenient when you only have four colors, but if you move to an instrument that allows bigger panels, manual comp becomes pretty cumbersome, and it ends up being easier and probably better to compensate using FlowJo.

Hope this helps, 

Max

Hi Max,

Thank you for your prompt reply. I'm only working on four colors now and I already find it cumbersome but still doable with proper understanding of the basics and utilizing currently available techniques and analysis softwares.

I think I have read a lot and thought I understand everything. However, once the actual experiment is being done, I'm starting to get lost especially when the actual results is far from my anticipated outcome. Thanks again, your answer gave me some relief and reassured me that somehow, I'm on the right track.

By the way, I believe that I just have to be able to view my events in FSC vs SSC plot, and at the same time should be on-scale with the FL-parameters, right?

Here's my general workflow:

1. Run unstained cells, make populations visible on scale in FSC vs SSC plot. Acquire.

2. Adjust PMT voltages for FL channels to place them on the LL corner (10^1 ~ 10^2).

3. Run my compensation beads, adjusting the voltages for FSC and SSC if necessary to make populations on-scale. But never adjust the PMT voltages for FL channels anymore. Acquire data per color.

4. Run samples, re-adjusting the voltages of FSC and SSC to that of the unstained cells on step 1. Acquire data.

5. Input FCS data into FlowJo and assign samples for Compensation Controls; apply gate accordingly as necessary.

6. Apply Compensation Matrix to samples.

I am still getting the hang of using FlowJo and its feature. Hopefully in the future, I would also be able to run 10-color flow cytometry experiments like you do. 

Thanks a lot again, your answer helped me understand compensation better.

Sincerely,

John

Hi John,

I agree with Max's comments completely.  One thing I would like to add:  The Calibur and CellQuest are analog systems.  As such, they apply compensation before the data is saved.  That makes it difficult, if not impossible, to change compensation later.  (Digital systems save uncompensated data and overlay linear/log designations as well as the compensation matrix when data is displayed.  The compensation matrix can then easily be altered, removed, or a new matrix substituted very easily).  To make full use of the compensation matrix calculated by FlowJo for analog files, it is important to use uncompensated fcs files.

Hi John

Both Lynda and Max already gave good answers. I would just like to add a small (maybe unrelated) but important step that I don't know if you are taking. Since you have a 4 stain flow, you should have a FMO also.

Thank you Lynda and Andre for these very helpful comments.

Our trainer before mentioned that, yes, the Calibur and CellQuest are still analog systems so it's like what I see on the graph is what I get. However, I was confused and sometimes run my samples and acquire uncompensated data, then run the comp controls the same way; I am too conscious with how the compcontrols look so I try to compensate them manually and acquire this data again but not doing the same in samples. Thank you for clearing this up Lynda.

Andre, I almost forgot about the FMO control, and I believe I have read in most articles that FMO is better than isotype controls in most cases. Thank you for this reminder.

Sincerely,

John

John

Just as an additional information. FMO and isotype control answer different things. I like using isotype control to verify "antibody stickness", i.e. if my blocking worked and if I washed the intracellular stain correctly. FMO will let you verify fluorescence spreading and garantee proper gating. Some people use isotype to gate, I don't like it because it can lead to false results. Besides that, most referees are old school and they always ask for isotype control so better be safe than sorry and have them both!

Happily do so, you won't regret.

Thanks a lot Andre and Jai, and thank you to everyone who shared their expertise to address my question. I learned a lot and will definitely apply it to my next experiment.

Hi John,

I would say it's preferably to calculate compensation in FlowJo, compared to do it manually.

You can absolutely acquire your samples without compensation, and then add a compensation matrix in FlowJo later, but your population of interest may not be visual without any compensation at all. Maybe a rough manual compensation first is a good ide. In FlowJo you can erase the rough compensation, change it, or make a new compensation matrix with the fcs-files of your single stained cells or beads. You must also include an unstained sample.

I have compared automatic compensation done by the software on the instrument, to the one calculated by FlowJo, and it does not even differ 1%. The cytometers I have used is FACSCanto II with Diva software, and FACSVerse with FACSuite software.

Good luck!

Kerstin

Hi Kerstin, thanks for your reply.

I will take note of you recommendations. About the rough manual compensation, this means I can still use the fcs data I acquired from samples where my compensation beads were acquired with manual compensation while the rest of the samples were acquired with no compensation?

If this is possible and valid for post-acquisition analysis, I will try do practice doing your suggestion with my existing data, and get something from those. At least, I might be able to salvage valid results from my previous experiment.

We only have FACSCalibur in the lab, still an analog version, and I have read that analog flow cytometers do not save all the information it acquires in fcs data: it only saves the actual data from the settings one have set. Thus, I might fcs data that don't have the uncompensated information in it for samples which I compensated before acquisition.

How I wish we have the digital flow cytometers like you have in your lab, I have read that these newer versions of flow cytometers records all the information it can get from the sample so one can do a more detailed analysis of fcs data.

Again, thanks for your time Kerstin, and goodluck with your current endeavours too. :)

Sincerely,

John

Hi John,

Sorry, I misled you. We previously had an Calibur in our lab. I tried some of the fcs-files acquired on it (with compensation), in FlowJo, and it seems like the compensation can't be changed in any way. I added a new matrix, but that did not change the appearance of the plots.

Maybe BD Biosciences or FlowJo can help?

Again, sorry about my, not helpful, advise.

Best regards,

Kerstin

Hi Kerstin,

Thank you for taking time to check your previous fcs files from Calibur with FlowJo compensation matrix. No worries, at least we're able to confirm that compensated Calibur-acquired fcs data cannot be edited in FlowJo in terms of compensation matrix. I think, as our BD Biosciences trainer said before, Calibur is still an analog version thus limiting the data saves upon acquisition unlike the newer versions of flow cytometry, the digital ones.

I thought of checking my own previous fcs data but I will not likely tell if I'm looking at the data right. I'm glad you were able to check and confirm, will definitely save me time. I will focus now on improving my run this week. I will just make the events on-scale in my plots since we know Calibur has some limitations like this.

Again, thank you very much for your help, all of your advices are helpful in a lot of ways. 

Sincerely,

John

Hello everyone,

On the subject already stated by John, I find myself in a very similar situation in a lab equipped only with FACSCalibur. I am currently setting up new panels and I would like to collect uncompensated data to later analyze in FloJo. My question is: do I need to collect the data using LINEAR instead of LOGARITHMIC transformation on all FL channels? My understanding is that this is an actual hardware controlled transformation (you can define it through the same control panel that adjusts voltage) and not just a reversible software function. Does everyone else that collects uncompensated data from a Calibur bother with that?

thank you all in advance,

Anne-Lise

Hi Anne, I am also wondering about that. Based on my previous runs, I tried combining Linear (FL3, 7AAD stain) with Log (FL1, FL2, FL4) settings when I acquired uncompensated data because based on my readings, viability markers via DNA staining should be on a linear scale. However, my bivariate graphs did not look good after compensation. Using FACSCalibur made me not forget that what you see on the graph is what you get after acquisition, thus, I just did the best what I could to adjust settings and see all events on my graph (following guidelines from a several literatures and online resources).

My suggestion is to use LOG consistently on all FL channels for immunophenotyping. And use LIN on channels assigned for proliferation or cell cycle analysis. For more information, you might find the link useful:

https://expertcytometry.com/the-difference-between-linear-and-log-displays-in-flow-cytometry/

Hopefully you could also share to us what worked best in your flow cytometry experiments. :)

Hi John, glad to see that someone still follows that question!

My concern comes from this (from FlowJo's documentation webpage)

" If FCS2.0 data was acquired on an analog instrument, the log converted in the machine will transform the data into log, so using a biexponential transform after acquisition will not work properly. These types of data are only useful if the file has been acquired in linear and compensated post acquisition in FlowJo. This is because these data files have typically been compensated and log transformed before the file was written, thereby eliminating the information in the file that makes the biexponential transformation process interesting or useful."

http://docs.flowjo.com/d2/graphs-and-gating/gw-transform-overview/gw-transform-procedure/

I consulted with BD's technical staff and they suggested this:

1. Keep the default "lin" for FSC and SSC and "log" for all other channels

2. Remove (if any) all threshold and compensation values

3. Adjust voltages appropriately in order to place the populations wherever you need them (ONLY VOLTAGES)

4. Switch to "lin" in the control panel for ALL channels

5. Collect your data (closing your eyes if necessary to avoid the dread of the uncompensated data)

6. Import your data to FlowJo and proceed accordingly

I will let you know how it goes after the first experiments

Hi Anne,

Thank you for sharing the BD technical staff's suggestion. An almost similar approach was suggested to me by BD representative here in our locality, except Step 4. But, this is actually a crucial step in a sense that a more raw data will be acquired by the FACS machine and Cell Quest via LIN but preserves the visual/graphical representation of the data set using the LIN (FSC & SSC) and LOG (FL1, FL2, FL3, FL4 channels) prior to collection of data. (I love the Step 5, closing my eyes to prevent seeing 'ugly' events and become skeptical of the protocol).

It seems like the key here is to acquire "rawest" data from FACS whenever one plans to use another software like FlowJo for post-acquisition analysis.

Too bad, I can do nothing about my previous data because they were already acquired via LOG on all FL channels; data acquired in LOG will somewhat interpreted by FlowJo as 'linear' data if one tries biexponential transformation.

I will try your procedure too, I can imagine now your protocol will work. :D

Thank you, I am curious and excited to know how your experiment will go.

Sincerely,

JP

Can you change the compensation in your experiment if say you had calculated compensation on the machine and it hadn't worked very well.. could you then run single stains and compensate in flowjo?

@Grace, using FACSCalibur, that would not be possible when you already set compensation settings prior to data acquisition. This is why a possible work around for this particular machine is being suggested. Based on the answers and experiences shared, if you would like to do post-acquisition analysis using 3rd party software like FlowJo, it would be better to run all your samples including the single stains that were prepared in similar batch at uncompensated at LIN scale on all channels; this is after you have a visual idea where you anticipate your events in LOG scale (FSC, SSC, FL channels). You may trying working on your idea as well, as long as it would not compromise the Rules of Compensation. I hope you'd let us know how it goes for you, too.

Dear John

I was kind of blown away when I saw that there were still some others having the same compensation issues with a FACSCalibur! I was wondering if the 6 steps shared by Anne worked for you? Do I really have to do step 4 before acquisition?

Thank you in advance for the valuable information shared in this topic.

Hi Milad Moloudizargari , thank you for your comment and I am glad you still find the information in this thread valuable. Unfortunately, I was not able to apply Anne's six-step approach because the project that I was working on was discontinued. I am still looking forward to try these steps the next time I'd have the opportunity to work on flow cytometry experiments again.

As for your second question, the step 4 would be the most critical part of this protocol as a work around for the analog set up of BD FACSCalibur. If you will not switch to LIN before you acquire, exporting your data to third-party flow cytometry analysis software won't be able to help you go back or trace your original data. You may think of it as having a raw data, without any adjustment made by the CellQuest software. Remember, using BD FACSCalibur, what we see on the computer interface is what we get.

I hope you could share your progress as well so we can further improve and help other researchers do their experiments well similar to this topic. Good luck and I wish the best for your work :)

Hi Anne-Lise De Lastic , I hope you are doing fine. I wonder if you could share some insights and your success story upon applying the 6 steps that you shared in this thread. Thank you in advance and my sincerest regards.