biocytex@biocytex.com (biocytex on 10/26/99 05:47:06 PM
To: cyto-inbox
<cytometry@flowcyt.cyto.purdue.edu>
cc:
Subject: Re: Statistics, scale and ratio? A case where K-S applies?
Dear Dr Boyce :
I am not sure K/S statistics is really needed is such a case. However, it
seems that quantitative
cytometry may be of help.
If I understand well, the aim of the proposed comparison is to determine
whether
different cell lines express different amounts of a given antigen /
receptor.
Am I right ?
>Yes that is correct. But I am looking intracellularly
In case you would not have easy access to flow cytometry, what else could
you do ?
A binding assay with radiolabeled antibody (Ab*) would be a classic
alternative.
Ab* binding is measured as counts per mn (cpm) per (let say) 10e6 cells.
Then, correcting
the test-Ab cpm with the background binding (e.g. cpm in excess unlabeled
Ab) ,
dividing by the specific activity of the Ab* and by the number of cells per
test gives
you the mean number of Ab molecules bound per cell (AB/C or ABC). That
number is
an absolute measurement and thus can be compared to a similar determination
made on the same cells an other day. AB/C may also be compared among
different
cells provided you stress that the unit is the whole cell and not a real
unit of cell
surface (for example
µm²). ( In case you really want to compare the results
as Ab
molecules per surface unit, then you have to assume the cell is a perfect
sphere and
measure its volume using the Coulter principle to determine the surface).
Measure AB/C on an absolute basis is typically what quantitative cytometry
can
easily do (and much more). As an example, the expression of endothelium-
associated cell-surface antigens has been compared on various cells of
endothelial
origin (see George F. et al. (1995), in: Schlossman SF et al. (eds),
Leucocyte Typing
V, pp1798-1801). This showed , for example, that the typical AB/C for
anti-endoglin MAbs (CD105)
were very different on HUVEC ( ~ 1,200,000 molecules/cell), on the cell
line Eahy 926
(~200, 000) and on the cell line ECV 304 (~20,000). Although the scatter
signals
were not exactly similar between cell types , this did not impede comparison
of the
calibrated mean fluorescence intensities.
As an other example, one can also compare the expression of an antigen
common to
cell types of very different size . DAF ( CD55 ) has logically very
different expression
levels on HUVEC (2 to 300,000 molecules/cell, same ref. as above, fig.1) and
on
human erythrocytes (4,800 +/- 1,400 , unpublished internal data). Although
these
measurements were made at very different periods (several years) using
different
calibrators and protocols (indirect IF with washings on cultured cells
versus a 2-step,
whole blood, no-wash quantitative assay), we can compare them. Both
experiments
(or series of) have used saturating doses of the same MAb and the
calibrators are
matching each other, although they have a different size and cover very
different
ranges of AB/C.
Once "normal" value ranges have been established for a given antigen, then
cell
(blood) samples can be compared for their levels of expression. As
previously
suggested (Ulrik Sprogoe-Jakobsen on oct. 12), the important issue is the
repeatability of the measurement on the same sample. Values falling out of a
range
of repeated values (e.g. mean +/- 2SD) can then be considered as
significantly
different.
Hope that helps.
Philippe Poncelet
BioCytex
Marseille, F
----- Message d'origine -----
De : Christopher S Boyce <csboyce@beckman.com>
À : Cytometry Mailing List <cytometry@flowcyt.cyto.purdue.edu>
Envoyé : samedi 23 octobre 1999 00:07
Objet : Statistics, scale and ratio? A case where K-S applies?
Hello Phillipe,
I appreciate your ideas very much. My questions are more hypothetical to try
and understand more about the proper way to set up a flow experiment from my
perspective. I am looking intracellularly for the anitgen(s) rather than cell
surface. I am specifically interested in the relationship between the two
antigens, and not so much the actual ABC because I am doing high throughput
screening of mabs. However, quantitative ABC will be the issue later in my
studies when desiring to know the level of expression. I do recognize
that flow is probably not the best format for measuring ABC, and the radiolabled
(cpm) approach would be useful. The ratios I describe are being used to try
and get a handle on the "normal" range of the relationship between the two
antigens against the permutations of cell lines and multiple mabs---before
quantification. Thanks so much for your references and experience with the
subject. I have seen the K-S stat used for things it probably shouldn't have
been used for, and I wanted to throw out a good example for everyone to "chew
on," and see what happens.
I intend to summarize everyone's responses and get them out later at a later
time.
My Regards,
Christopher
This archive was generated by hypermail 2b29 : Wed Apr 03 2002 - 11:54:10 EST