ABCDFITTER was developed originally to run in batch mode under OPEN_VMS. In its current
incarnation, the DOS/Win95 version, ABCDFITTER.EXE, reads in a control file that
contains all the initial set up conditions: number of points, beginning and ending
radial positions, etc , the initial parameter guesses, which parameters to hold
constant and which to float during the fit. The program runs for a fixed maximum
number of iterations or until the parameters converge to within a specified tolerance.
ABCDFitter is still under development: it needs a more friendly "graphical user interface"
front end. This will appear in the future. It does the work you need now; beauty will
have to wait. It writes 2 output files. One is a file that contains the values of
the fitted parameters at each new minimum and the second file is a much more extensive
"log" file that contains a trace of the whole fitting procedure.
It is optimized for the 2 step binding system:
A + B = C K1
C + B = D K2
If you want to fit the simpler 1 step binding reaction:
A + B = C K1
Set the value of K2 to 1 x 10^-20
Set SD = SC and hold these variables (2 and 4) during the fit i.e. 01010000Š
The program assumes (1-vbarxrho)= 0.275 and that
An example control file for a 3 cell global fit to A + B = C to interference
data follows .
It can be run with the comments left in for your future reference; just substitute
your own values for the parameters and guesses.
------------------------- cut here X---------------------------
N do not do a monte carlo analysis [see example below for monte carlo]
3 3 cells are to be fit globally
200 x-axis number of points for claverie routine | Use these
1.0 DT time increment for claverie routine | values for now
CELL_1_LIST.DAT list containing names of files from cell 1.
fringes concentration (could be "mg/ml" or "absorbance")
5.8687 meniscus position ^ if "absorbance", next 2 lines are
5.90 first point to include. Molar ext'n coeff's of A and B. (see below}
7.2 bottom position
7.0 last point to include in fit
n no do not align scans above the meniscus
CELL_2_LIST.DAT etcŠ
fringes
5.8806
5.90
7.2
7.0
n
CELL_3_LIST.DAT
fringes
5.8971
5.92
7.2
7.0
n no do not align above the meniscus
y yes write dcdt-out for cell 1
y yes write dcdt-out for cell 2
y yes write dcdt-out for cell 3
1 1 = minimize RMS residual; 2=minmize absolute value of residuals
21.e3 MW of A |
18.e3 MW of B | known parameters
1.56 SA |
1.70 SB |
3.19 SC guess|
3.19 SD guess| global fitting parameters 1-4
4.0e5 K1 guess|
1.0e-20 K2 guess| this forces the system to A + B = C
0.0555382 CA1 loading Concentration (mg/ml) of A cell 1.
0.0531800 CB1 loading Concentration (mg/ml) of B cell 1.
0.159823 CA2 etcŠ
0.155718 CB2
0.435684 CA3
0.522352 CB3 | CA1 - CB3 are local fitting parameters 5-10
0101000000 hold variables 2 and 4 constant, e.g. SD and K2
500 maximum number of iterations
------------------- cut here X------------------------
This file should be self explanatory. The data sets must be plotted separately in
another plotting program to determine the meniscus and base position and the subset
range between those points to be used in the fit. A future, more user friendly,
version will allow graphical choice of the meniscus, base and end points of the
data. It is sufficient in the base region to cut the data off before the "hinge point"
and to set the cell bottom a little bit beyond the actual cell bottom [i.e. = like a Faxen
type infinite cell approximation ]. This eliminates the need to know accurately the
cell bottom position which would otherwise be necessary to get good fits at the base.
An accurate meniscus position is required, however.
ABSORBANCE DATA
Here is an example of a control file from a 6 cell fit to absorbance data for the system
A + B = C K1
C + B = D K2
--------------------- cut here X---------------------------------
n
6
200
1.0
vk582.dat file list for cell #1
absor this is absorbance data
7.770E4 extinction coefficient of A
1.20E5 extinction coefficient of B
5.939
5.98
7.2
7.1
n
vk581.dat file list for cell #2
absor
7.770E4
1.20E5
5.985
6.02
7.2
7.1
n
vk221.dat etc...
absor
7.770E4
1.20E5
5.930
5.96
7.2
7.1
n
vk222.dat
absor
7.770E4
1.20E5
5.924
5.95
7.2
7.1
n
vk223.dat
absor
7.770E4
1.20E5
5.920
5.95
7.2
7.1
n
vk583.dat
absor
7.770e4
1.20E5
5.967
6.00
7.2
7.1
N
Y 1 keep track of the number of responses on the next n lines
Y 2 where n= number of cells.
Y 3
Y 4
Y 5
y 6 end
1 RMS being minimized (=2 if abs value of residuals are to be minimized]
95838.05 Ma
100000.00 Mb
3.284 etcŠ as in previous example
3.617
6.114 1
7.205 2
4.41e5 3
4.75e7 4
0.38 5
0.10 6
0.29 7
0.10 8 <---- variables 1-16
0.19 9
0.10 10
0.19 11
0.20 12
0.19 13
0.40 14
0.19 15
0.80 16
0000000000000000 = let all 16 variables float
500 = total number of iterations
------------------------------ cut here X--------------------------------
MONTE CARLO ANALYSIS
To do a monte carlo analysis after converging to a set of parameters, generate n sets
(where n is equal to the number of cells combined in the global fit) of perfect data
using the resultant parameters obtained in the fit.
The top few lines of the command file are changed to read :
Y yes, we're doing a monte carlo analysis
0.00 mean of the noise to be added to the ideal data
0.005 RMS noise to be added to each trace. Each trace gets unique gaussian noise.
N No,[This lets the program choose the seed for the random number generator]
N do not write out data files containing the added noise. Y if you want them.
6 number of cells for global fitting
and so on ... the rest of the file is the same as above.
This command file may be used repeatedly to generate each new fit with a different
set of random noise for every "data" file in the synthetic data set.
Repeating this 10 times by launching 10 copies of abcdfitter specifying this command
file but specifying a different output file name each time will generate 10 fits
whose returned parameters can be analyzed to obtain estimates of the standard
deviations of the fitted parameters for this model. This is an abbreviated monte
carlo analysis and is sufficient in most cases to give you an idea of how well
determined each of the fitted parameters is.
The time to do 10 fits for an abbreviated monte carlo analysis for 6 cells can
take up to 24 hrs. on a 300 Mhz P-II.
ABCDFitter will run in the background if launched from Win95.
Please contact me if you have any difficulties or concerns: stafford@bbri.org
Walter Stafford
14-APRIL-1999
last minor update 16-NOV-1999