Normalizing data

From: Jim Zanghi (zanghi@merle.acns.nwu.edu)
Date: Thu Jan 23 1997 - 20:54:19 EST


I'm going to word my problem in a generalized way, since I've seen this
come up in several diffent types of experiments:

I have examined two cell surface antigens, A and B, using single color
(FITC), indirect antibody staining and flow cytometry.    The cells are a
stable line that is 100% positive for both A and B under all circumstances.
I wish to quantitate changes in A and B under various treatments by
examining the mean fluorescence intensity of flow samples.

For Treatment X,  A goes up slightly and B goes down dramatically compared
to no treatment (control cultures stained with A and B antibodies).
Changes in both A and B are both  reproducible and statistically
significant.   However, the autofluorescence signal (cells stained with
buffer without any antibody) also goes up slightly for cells under
Treatment X compared to no treatment.   (None of the isotype controls are
affected under any condition or under any circumstance and are always
identical to the autofluorescence signal.)   I have the voltage set such
that the autofluorescence signal is trivial compared to the signal of
either A & B, so I didn't bother subtracting it out.  Instead, I tried
normalizing all my data to the autofluorescence signal by taking the ratio
of Treatment X stained cells (XA or XB) over Treatment X unstained cells
(Xauto) and compared this (XA/Xauto or XB/Xauto) to untreated control cells
(CA/Cauto or CB/Cauto).   By doing this, I get a different interpretation
of my results.   The increase in A is very similar to the increase in
autofluorescence, or maybe even less, so that with the normalization, I
find that Treatment X has no effect on A, or possibly a decrease in A,
compared to no treatment.   The effect of Treatment X on B is even more
pronounced after normalization.

How do I interpret my results?  What is the significance of a change in a
cell's autofluorescence?  I notice that by changing the voltage on the
instrument, the ratios of A/auto and B/auto both remain constant, but does
this necessarily justify the normalization procedure to cancel out
biological-based changes in a cell's autofluorescence?    In this
circumstance, it appears justified because the changes in A and auto are
directly proportional.  Or might this just be coincidental?  The
proportional changes in A and auto are very similar over both time and
concentration of X.

Should changes in forward scatter be proportional as well?

Thanks,
Jim

--
James A. Zanghi
Dept. of Chemical Engineering
Northwestern University
Evanston, IL



This archive was generated by hypermail 2b29 : Wed Apr 03 2002 - 11:49:19 EST