 |
 |
 |
 |
Thank you for your reply to our query about our poor reproducibility (up to
2.5 fold discrepancy) between replicate FACS titrations of antibodies. (Its
the only reply we've received to date.)
We noticed several differences between your methods and ours.
First, your use of 2-fold instead of 3- or 4-fold dilution steps
may be important. Second, you fit your data by the Lineweaver-Burk (LB)
method, while we fit ours directly to one-site binding (OSB) kinetics.
We have fitted one of our worst pairs of curves by both methods. The results
suggest that you might want to switch to the method we are using. As you
point out, LB fits are excessively dependent on the least reliable points
(lowest concentrations of Ab). As shown below, dropping the lowest intensity
point makes a large difference in the values obtained. In contrast, one-site
binding curves are relatively insensitive to dropping this point. Also, LB
compresses the high end of the curve which contributes to losing much of the
information there; this is not so in the OSB method. The equations are:
I = fluorescence intensity
Imax = I at Ab saturation
C = concentration of antibody
Ch = C giving half-saturation
* signifies multiplication
LB: 1/I = (Ch/Imax*C) + 1/Imax
(A plot of 1/I vs. 1/C is a straight line)
OSB: I = (C*Imax)/(Ch + C)
(We use a log log plot of I vs. C, giving an asymptotic saturation curve)
The results are (I for lowest C dropped for 3 point fits)
Curve 1 Replicate
(poor fit) (good fit)
------------- -------------
Ch Imax Ch Imax
LB (4 data points) .0468 525 .0103 323
LB (3 data points) .0611 645 .0159 392
OSB (4 data points) .0229 388 .0123 361
OSB (3 data points) .0224 386 .0123 361
Notice that
(i) LB is much more sensitive to the lowest C data point (which nevertheless
has a signal several times the autofluorescence level);
(ii) LB gives poorer agreement between the replicate curves;
(iii) LB results agree poorly with OSB results when there is a poor fit.
Incidentally, we use a marvelous PC scientific graphing program to do the
curve fits with the undistinguished name of "Fig P". It makes fitting very
easy, and you can put in any equation you like! It is available from Biosoft
(314-524-8029) for about $500. Expensive, but worth it. It makes publication
quality graphs, and is easy enough to use that everyone in my lab uses it
in preference to hand plotting!
I'd like to get 3 things from you:
1. The raw data (I and C values) for a couple of the curves you did -- one
in good agreement and one in poor agreement. We'll put them into Fig P
and try both methods.
2. Your FAX number so I can FAX you some graphs.
3. Details of you dilution method. We still need to make our replicates as
reproducible as yours. We use Biotip pipets (whose accuracy and precision we
have checked by weighing deliveries) and serially carry 25 or 50 microliters
from tube to tube containing appropriate diluent volume (e.g. 75 or 150 ul).
Usually we mix only by pipetting in and out, and do not change the tip.
(I'm sending this to the cytometry list so you'll get an extra copy.)
--
/* - - - - - - - - - - - - - - - - - From - - - - - - - - - - - - - - - - - -
Eric Martz emartz@titan.ucc.umass.edu Professor of Immunology
Voice 413-545-2325, FAX 413-545-1578
Morrill IVN Rm 203, Dept Microbiol, Univ Mass, Amherst MA 01003 USA
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
|
|
|
|