Isotypes: "Intelligent Design" for science
Folz-Donahue, Kathryn E.
Kathryn_Folz-Donahue at dfci.harvard.edu
Wed Oct 26 12:03:26 EST 2005
That is the best analysis of the problem & solution I have seen so far. Bravo.
I am printing it out and putting it on the wall.
Flow Cytometry Core Facility
Dana-Farber Cancer Institute
From: Uriel TK [mailto:utk1 at 013.net]
Sent: Friday, October 21, 2005 8:23 AM
Subject: Re: Isotypes: "Intelligent Design" for science
I'd like to comment that I completely agree with the mentioned problems of
isotype controls and their usage. I guess most users think they represent
something they do not or ignore their shortcomings, and I include myself in that
last category maybe too often. But I disagree with the recommendation to throw
them away. A more practical recommendation would be to be critical of results
and procedures, including controls used, without taking the isotype control as a
dogma but rather as another control, imperfect as it is. Certainly not as the
only control! Most 1st rate cytometry laboratories and research groups (and most
of the rest of us!) use them and have used ICs. Something is telling us through
the years to use them and I think it is more than just inertia. I think it is
the need for an "unspecific binding" control. Many times, even if you block Ig
and add BSA and etc there is a certain fluorescence increase in cells "negative"
for the mAb used. How do you deal with that? Is that low expression or
unspecific staining? We have to deal with that often and I can't seem to find a
good enough answer. The problem is worse when using 2ry Abs, where without the
addition of PROPER (isotype) controls it can be very hard or impossible to know
what is going on. Even more so with intracellular staining. So the IC come to
give a partial answer to an important question which has no good-and-easy
answer. What I'm saying is that for the money and effort, the information
obtained is better than none, and the more rigorous you are on their use the
more you obtain from them. This IC-information should include its shortcomings
of course. And yet, even if we had perfect ICs (if suppliers gave us Ab
concentration and F/P ratios), in all sincerity, who is going to use a matching
IC tube for every test tube?
On this subject of the need for discriminating between low expression and
unspecific binding, which is the main use of ICs-as-escape-goats, I think an
important note has to be made. I first consciously realized this when I heard
Nicole Baumgarth in her video in the "Cytometry 8" CD. When asked what to do
with discriminating low expression from artifacts (in that context it had more
to do with multi-color compensation) she said (more or less) that "if you can't
say then probably you can't say and should use another assay to test for that,
that flow cytometry is not the answer for all and if the call is too tight then
maybe it is really too tight to call it". Since then I sleep better. If the
answer is important then sort and do a western or use a system with a better
signal to noise ratio (less colors and 2step amplification for example), but
don't be fooled into thinking that you CAN say if you CAN'T say within the
confines of the experimental system being used. It is tough since it means we
have to do more work, but for me it gave a lot of peace of mind since it
released me from the burden of having to reach a decision when I am not
completely sure. Now I know that flow maybe won't give me the answer and leave
it at the uncomfortable but sound point of "not really known", even if I believe
(in the intelligent design way of believing) it is that way or the other. The
brightest part is that the answers I got when going the stringent way were
better (more categorical) than what I got with FC alone. And lets not forget
that the mean and median and other statistical measures are still there to use
when %-positive doesn't cut it.
In this respect I want to add that the too-common use of setting a marker on the
last 1% of the IC and saying "all the rest = % positive" should be banned. In
general my view is that unless the separation between positive and negative
peaks is a LOW valley, say 10% or less of max peak height, separating with a
marker is delusion. It hard to let go of that custom but that is biological
life! Maybe I'm being too hard, but actually there is an easy solution: if the
software makers added curve fitting algorithms to phentypic expression
distributions we could let go of the markers AND ICs AND get population numbers.
These algorithms are already there in proliferation models and the porting
shouldn't be a problem. If you are reading, please hear our plight!
I think in the end it comes down to us. Will we look at our friend's result and
say "this is not the correct way to use ICs?" or "you need another control"?
Will those of you reviewing articles ask for revisions due to the sole use of
ICs as controls? The same can be said, by the way, on the use of statistic
tests: How many of us use t-test for linear data which is non-parametric? Worse,
how will a reviewer relate to the article if it uses a non-parametric test he
doesn't know instead of t-test "as is the norm'?
The Laboratory for Cellular and Molecular Immunology
The Hebrew University - Hadassah Medical Organization
Jerusalem - ISRAEL
"God does not care about our mathematical difficulties. He integrates
-------------- next part --------------
HTML attachment scrubbed and removed
More information about the Cytometry