RE: newbie help with software compensation

From: Mario Roederer (roederer@drmr.com)
Date: Wed Jun 19 2002 - 15:47:00 EST


>In principle I agree with you that the computer is more accurate in
>performing compensations on the more complex experiments.  I am merely
>asking this so you can put everyone's (at least my) first question at ease.
>Are the computers always right?  Have you found any potential bugs in the
>"system" where the computer (and I am not assuming any particular software)
>is incorrect because of some unforseen influences in the input data?  I know
>this is vague, but while I trust in the programmers' abilities, you are
>actually doing the most complex experiments with more colors and may have
>found some caveats for which we must all be aware.
>
>Randy T. Fischer

Randy,

the question you ask is not so simple to answer, and would take much
more time than readers of this list would wish to spend.  I refer you
to my web site and recent paper on compensation for a variety of
caveats, unseen influences, etc.

First of all, even without personal experience with most programs, I
have a very high level of confidence in the ability of these programs
to compensate properly when given the right input data.  After all,
the algorithms behind compensation aren't rocket science; they are
very simple linear algebra... and, once vetted on a few samples, are
likely to be accurate.  And note that if you had perfect compensation
controls, then the process of compensation could be perfect.

So what are the major things to look for that might lead to
"inaccurate" compensation?

The most important one is having good compensation controls.  If any
of your stained samples are brighter (in any given channel) that the
corresponding compensation control, then you will more than likely
have a problem.  This is because the error in estimating proper
compensation is related to how bright a control sample is--in
general, you can assume that cells brighter than your comp control
are probably not compensated correctly!  But this is easily solved:
use the brightest stain you have as your comp control.

Also make sure that you are gating properly when identifying the
compensation controls.  For example, if you use FITC CD8 as the comp
control, don't use an unstained ungated PBMC sample as the "negative"
control, because it contains monocytes that have much higher
autofluorescence that the stained "positive" CD8 T cells.  Make sure
that for any given control, the "positive" and "negative" (or bright
and dull) cells have the exact same autofluorescence characteristics
before staining.

Finally, if you are using tandem dyes (Cy5PE, Cy7PE, TRPE, etc.),
then you will likely need a different compensation setting for each
different reagent.

There are other "caveats" and problems, but they are generally much
smaller in magnitude than the error from having poor comp controls.
For example, most log amplifiers are notoriously
inaccurate--therefore, if you use software to compensate the data
(and the software must assume that the log amps are good), then you
can run into problems where one range of brightness is properly
compensated, but cells above or below that can be under- or
over-compensated!  For cytometers where log amps are used (a vast
majority of extant machines), I recommend that you use "partial"
compensation on the machine (i.e., guarantee undercompensation), then
collect all of your single stained comp controls, and then use
software to finish off the compensation.  The major advantage of this
is that you are pretty safe using the same settings day in and day
out, and the software ends up correcting for the day-to-day variation
in instrument sensitivity.

By the way, another common error is to use the same comp settings day
in and day out.  STOP DOING THAT!  I know it's convenient, but it's
wrong.  You should have comp controls for every experiment and set
compensation every time.

mr



This archive was generated by hypermail 2b29 : Sun Jan 05 2003 - 19:26:13 EST