[Cytometry] Compensation with liver cells

Nebe-Von-Caron, G g.nebe-von-caron at invmed.com
Mon Jul 13 19:55:05 EDT 2009


As the compensation is not based on the number of photons detected in he
respective channels but on the output voltage generated by those
photons, changes of the amplification of those channels change the
compensation values. So if for a fictitious example with a 1mV offset
PMT1 turns 1000 photons into an output voltage of 1001mV and the second
PMT receives 100 photons through its filter by spectral overlap which
would create a 101mV signal. A 10% compensation from PMT1 to PMT 2 would
subtract the 10% of 1001mV making the output of PMT2 to 1mV. If one
suddenly reduces the amplification of detector 1 by a factor of 4 to
compensate for the high autofluorescence the 1000 photons from your
antibody would now give rise to 251mV and therefore the 10% compensation
would only subtract 25.1mV from the second PMT leaving 76mV signal. Now
if one were to reduce the amplification of the second detector by a for
example a factor of 2 to leave the autofluorescent cells occupying the
first decade in both channels  detector 2 would respond to the 100
photons from the fluorochrome with 51mV signal which still leaves 26mV
after compensation.

If the amplification can be changed on all fluorescent detectors by the
same amount one way to simplify the procedure is to tune down the laser.
On the other hand this might be a good argument for the Accuri approach
not to change the detector voltage but to scale the data to the required
observation range. Alternatively if you can change your channel
amplification by a known factor (e.g. you know your detector response
curve) you can recalculate the compensation matrix. The key take home
message is that a change in detector amplification changes your
compensation settings. 

Now in theory there is no difference between theory and practice but in
practice there is

Whilst the emission profiles of the fluorochromes are relatively
constant apart from changes due to pH, energy transfer, bleaching
effects, conjugation instabilities, the position of Venus or Jupiter....

the autofluorescence values of cells may not be so. Some
autofluorescence is a consequence of cell age, stress, stimulation...
and thus may vary significantly between control and test samples so one
needs to include autofluorescence controls for several samples and check
how they change over the course of the experiment. 


Hope that adds to the confusion


Gerhard



Gerhard Nebe-von-Caron 
Research Scientist and Biomedical Engineer 
Inverness Medical International
Priory Business Park 
Bedford, MK44 3UP, UK 
Mob +44(0)7792-116609
Tel +44(0)1234-835474 
Fax +44(0)1234-835002 

mailto:g.nebe-von-caron at invmed.com 
 

-----Original Message-----
From: cytometry-bounces at lists.purdue.edu
[mailto:cytometry-bounces at lists.purdue.edu] On Behalf Of Mario Roederer
Sent: 13 July 2009 13:31
To: cytometry at lists.purdue.edu
Subject: Re: [Cytometry] Compensation with liver cells

For those of you who are math-averse, or compensation-averse, you can
skip this entire email!  However, the only way to put this issue to rest
will be to go in deteail through the compensation math.

Simon, you are correct that, on older analog instruments, when the
negative bead population is on the lower axis, the median will be
reported as 1, as  you note.  And, this median value is incorrectly
high.  Presumably, it should be something like.... 0.3 or 0.1.  Let's
assume, in the most extreme case, that the value should have been 0.  
(Since it can't be negative, this is as much inaccuracy as is possible).

Let's further assume that you have turned down the voltages to get your
cells on-scale (i.e., between 10^3 and 10^4), and that you have
correctly used beads whose positive fluorescence is at least as bright
as your cells.  Therefore, your positive bead median fluorescence is (at
least) 10^3 (1000).  We'll use 1000 as the "worst case"  
possibility.  (This is the worst case, because if it is lower than this,
you need to turn up the PMT voltage and then hopefully your negatives
will come up off the axis as well).

The compensation coefficient is the ratio of the DIFFERENCE of the MFI's
(of the spillover channel and the primary channel).  Thus, in this
channel, the difference will be computed as "positive - negative MFI" =
"1000 - 1", which, as you note, is slightly incorrect (it should be, in
our extreme case, "1000 - 0").  Assume that the difference in MFI of the
positive and negative bead population in the spillover channel is "dS"
(delta-spillover).

Thus, the compensation coefficient will be computed as

ds / (1000 - 1) = dS / 999.

instead of the exact

ds / (1000 - 0) = dS / 1000.

You will note that the difference in these values is one part in 1000.
Thus, if the correct spillover value was supposed to be 15.0% (i.e., dS
= 150), by having the beads on the axis, it now comes out to 15.015%.

There is no meaningful difference between these two compensation values.
In fact, the precision on the compensation measurement is far less than
this difference -- i.e., the typical variance on the computation of a
compensation coefficient will be on the order of several percent, and,
unless you are collecting tens of thousands of beads for each
compensation control, may well be 5-10% or more.  (This error is
primarily driven by the imprecision of the median of the positive,
bright beads).  Since this is one to two orders of magnitude greater
than the inaccuracy introduced by pegging the negative beads on the
axis, the contribution of the latter is negligible.

The bottom line is that because the compensation coefficients are
computed based on the DIFFERENCE in MFI's (and not the RATIO of MFI's),
small absolute errors in the position of the negative control become
irrelevant if the positive controls are nice and bright.  The error in
the compensation coefficient is the sum of the absolute errors in the
MFI's of both the negative and the positive control; the latter has an
inherently MUCH larger absolute error than the former.

mr


(Incidentally, if you peg the negative beads in the spillover channel as
well, and there is very low spillover (dS is <10), then the error in the
compensation coefficient might be much larger.  However, as long as your
compensation control is at least as bright as your cells (this should be
your mantra), then even this larger error will not affect your
measurement, because the absolute magnitude of this error on the cells
will be much less than the variance in the autofluorescence of the
cells.)




On Jul 13, 2009, at 5:55 AM, SIMON MONARD wrote:

> Ian
> All the instruments in my lab are NOT digital. Probably most 
> instruments in the world are not digital. Granted you could use comp 
> beads on digital machines when looking at large cells. However, on a 
> Calibur, as soon as 50% of the blank beads are in the first channel 
> the median will be 1. Continue to reduce the voltages and it will 
> remain 1. Likewise your positive beads will have a median of 1 when 
> 50% are in the first channel and will continue to be 1 as you increase

> the compensation. So, the compensation will apparently be correct for 
> a wide range of PMT voltages and compensation settings. which ever way

> you visualize and calculate compensation this will not change the fact

> you have all those uninformative ones.
> I am suggesting an inexpensive solution that works with all platforms 
> that can be used with real time or software compensation.
>
> Simon
>
>  Quoting Ian Dimmick <ian.dimmick at newcastle.ac.uk>:
>
>> Beads can not go off scale , the relative negative events will 
>> allways be visualisable to whatever degree, it has been my experience

>> that you can always compensate on the calibur using these beads , you

>> would stop your compensation when the median of the bright was the 
>> same as the negative , or if you wish you could import the Calibur 
>> data into diva and use the bi exponential scale to visualise a little

>> better the negative compensation beads , although still using the 
>> median.I did not ignore Simon's comment regarding the 
>> autofluorescence , if you re read my mail I stated that 
>> autofluorescence is taken into account in the inverted matrix 
>> calculation , whatever value it has , you defend Simon but I do not 
>> feel his position needs to be defended , we all have our ways of 
>> doing things and we all have our views on "the best" way to do 
>> things, this is why this forum thrives, in many scenarios we are all 
>> helped by people offering their comments on personal experience , and

>> very ofte!
>> n those views will differ , primarily because flow cytometry ,cells ,

>> filters , lasers....................are not perfect , and very far  
>> from standard, sometimes even within the same instrument platform ,  
>> so please if you want to make comments then make them in context of  
>> your own experience , do not miss quote, I felt that the whole  
>> discussion was starting to have the effect of confusing rather than  
>> helping , I did not aim that at Simon , but the whole thread in  
>> general , however if Simon felt that my comment was aimed at his  
>> mail the I appolgise to him .
>> I do however stand by my comments
>>
>>
>> Ian Dimmick
>> Flow Cytometry Core Facility Manager
>> Institute of Human Genetics
>> Bioscience Centre
>> International Centre for life
>> Newcastle Upon Tyne
>> NE1 3BZ
>> UK
>> Ian.Dimmick at ncl.ac.uk
>> Tel 0044 191 2418831
>> Fax 0044 191 2418666
>> (mob) 0044 7970344823
>> http://www.ncl.ac.uk/ihg
>> ________________________________________
>> From: cytometry-bounces at lists.purdue.edu
>> [cytometry-bounces at lists.purdue.edu] On Behalf Of Ray Hicks 
>> [rhicks at cytekdev.com]
>> Sent: 11 July 2009 16:07
>> To: 'Mario Roederer'; cytometry at lists.purdue.edu
>> Subject: Re: [Cytometry] Compensation with liver cells
>>
>> So Mario and Ian,
>>
>> On an analogue machine, say a Calibur, if you wish to set 
>> compensation for bright cells, you'd set your PMT voltages so that 
>> the negative cells would be on scale, then you pop on your 
>> compensation beads, the unlabelled comp beads are off scale or 
>> clipped to the origin of the graph, the stained comp beads are on 
>> scale, you start compensating them so that the spillover fluorescence

>> is equivalent to that on the unstained bead, you hit a pitfall, in 
>> that you don't know when to stop.  Whether you use mean or median or 
>> magic, you're not going to know when you've applied enough 
>> compensation, since you don't know where the negatives lie.  Using 
>> compensation beads that have enough "Autofluorescence" to be on scale

>> at the same time as the stained ones is pretty much essential I'd say

>> under this circumstance - one that Simon described and which you've 
>> both ignored in your answers.
>>
>> Care would probably have to be taken that the bead autofluorescence 
>> was neutral over the wavelengths investigated.
>>
>> I don't think Simon was confusing any _issues_ Ian
>>
>> Ray
>>
>> Ray Hicks
>>
>> http://www.cytekdev.com
>>
>> Tel: +44 (0)208 1337 968  Fax: +44(0)208 5889 004  Skype:  
>> ray.hicks.cytek
>>
>> Cytek Development Europe, Unit 8 The Maltings, Millfield, Cottenham, 
>> Cambridge CB24 8RE. UK
>>
>>
>>
>> -----Original Message-----
>> From: cytometry-bounces at lists.purdue.edu
>> [mailto:cytometry-bounces at lists.purdue.edu] On Behalf Of Mario 
>> Roederer
>> Sent: 10 July 2009 20:43
>> To: cytometry at lists.purdue.edu
>> Subject: Re: [Cytometry] Compensation with liver cells
>>
>> As I wrote, as long as your positively-stained beads are at least as 
>> bright as your stained cells, they are a perfectly good compensation 
>> control.  In today's instrumentation, it's irrelevant if the negative
>> (unstained) beads are "not onscale".  With modern digital 
>> electronics, it's not even possible, since all signals are collected 
>> in linear, for which even zero measured fluorescence is still 
>> onscale.
>>
>> So in fact, the background fluorescence of compensation beads is not 
>> relevant, and there is no need to match it to the cells. Don't worry 
>> if the beads are pegged at the bottom when you run on the voltages 
>> optimized for your cells.
>>
>> So.... In summary:  No!
>>
>> mr
>>
>> On Jul 10, 2009, at 12:48 PM, SIMON MONARD wrote:
>>
>>> No...?
>>>
>>> If you are looking at larger cells, ESCs, Macs, hepatocytes etc you 
>>> will find that setting your voltages on the negative cells the beads

>>> will be offscreen. Bump up the voltages enough to get the beads on 
>>> screen and you may find you have little dynamic range left to 
>>> measure fluorescence. You can find that your negative cells are two 
>>> logs higher than blank comp beads. If you have an older machine with

>>> 4 log decades BD comp beads are useless if you work with such cells.

>>> The BD beads have clearly been selected because they have a similar 
>>> background to lymphocytes. Larger beads have a higher background but

>>> also capture more antibody so they can be a lot brighter than the 
>>> commercial beads. I'm not suggesting adding background fluorescence 
>>> to blank beads I'm suggesting selecting beads that better match your

>>> cells and coupling them to anti-rat Ig or whatever. 6 micron 
>>> polycarbonate beads are a much better choice for larger cells.
>>> In summary...Yes!
>>>
>>> Simon
>>>
>>>
>>> Quoting Mario Roederer <roederer at drmr.com>:
>>>
>>>> No... antibody capture beads are not designed with a particular 
>>>> cell subset in mind.  Any compensation bead will work for any cell 
>>>> type, as long as the positive beads are brighter (or at least as 
>>>> bright) as your stained cells.
>>>>
>>>> Adding background to your beads (or using beads with more
>>>> background)
>>>> will only decrease the precision (although not the accuracy) of 
>>>> your compensation.
>>>>
>>>> mr
>>>>
>>>>
>>>> On Jul 10, 2009, at 5:24 AM, SIMON MONARD wrote:
>>>>
>>>>> You will likely find them unusable as the commercial ones are
>>>>> designed
>>>>> for use with lymphocytes. If you are planning on doing much FACS
>>>>> then
>>>>> I suggest you make your own capture beads using beads with a
>>>>> background fluorescence comparable to your unstained cells
>>>>
>>>> _______________________________________________
>>>> Cytometry mailing list
>>>> Cytometry at lists.purdue.edu
>>>> https://lists.purdue.edu/mailman/listinfo/cytometry
>>>>
>>>>
>>>
>>>
>>>
>>> Simon Monard
>>> FACS Facility Manager
>>> MRC Centre for Regenerative Medicine
>>> Institute for Stem Cell Research
>>> University of Edinburgh
>>> Roger Land Building
>>> West Mains Road
>>> Edinburgh
>>> EH9 3JQ
>>>
>>> Tel. Lab 0131 6505876
>>> Tel Office 0131 6517265
>>>
>>> --
>>> The University of Edinburgh is a charitable body, registered in
>>> Scotland, with registration number SC005336.
>>>
>>>
>>> _______________________________________________
>>> Cytometry mailing list
>>> Cytometry at lists.purdue.edu
>>> https://lists.purdue.edu/mailman/listinfo/cytometry
>>
>> _______________________________________________
>> Cytometry mailing list
>> Cytometry at lists.purdue.edu
>> https://lists.purdue.edu/mailman/listinfo/cytometry
>>
>>
>> _______________________________________________
>> Cytometry mailing list
>> Cytometry at lists.purdue.edu
>> https://lists.purdue.edu/mailman/listinfo/cytometry
>> _______________________________________________
>> Cytometry mailing list
>> Cytometry at lists.purdue.edu
>> https://lists.purdue.edu/mailman/listinfo/cytometry
>>
>>
>
>
>
> Simon Monard
> FACS Facility Manager
> MRC Centre for Regenerative Medicine
> Institute for Stem Cell Research
> University of Edinburgh
> Roger Land Building
> West Mains Road
> Edinburgh
> EH9 3JQ
>
> Tel. Lab 0131 6505876
> Tel Office 0131 6517265
>
> -- 
> The University of Edinburgh is a charitable body, registered in
> Scotland, with registration number SC005336.
>
>
> _______________________________________________
> Cytometry mailing list
> Cytometry at lists.purdue.edu
> https://lists.purdue.edu/mailman/listinfo/cytometry

_______________________________________________
Cytometry mailing list
Cytometry at lists.purdue.edu
https://lists.purdue.edu/mailman/listinfo/cytometry



More information about the Cytometry mailing list