Nonlinear regression with residuals

Petr Kuzmic pkuzmic at biokin.com
Tue Nov 25 11:22:04 EST 1997


Matthew Parker wrote:
> 
>         Hi! Does anyone know of a curve-fitting program which can do
> nonlinear regressions to equations which the user specifies, and can
> give a plot of the residuals?


Try program DYNAFIT from http://www.biokin.com/ --> Software.  It is
suitable for the nonlinear regression analysis of chemical and
biochemical data (kinetics and complex equilibria).

The program not only generates residual plots (in the PostScript format,
so you can just paste them into a word-processing document), but also
the "user specified" equations are non-mathematical.  For example the
user specified equations for the binding of two dimeric protein
molecules (monomer ... M, dimer protein .... P) to one molecule of DNA
would be written by the following stoichiometric equations:

[mechanism]
   M + M <===> P           :   Kd    dissoc.
   P + DNA <===> P.DNA     :   Kd1   dissoc.
   P.DNA + P <===> P2.DNA  :   Kd2   dissoc.

Of course the names of (bio)chemical species and the labels  for rate-
or equilibrium constants are entirely arbitrary.

The main advantage of DYNAFIT over other programs (PRISM, ORIGIN,
SIGMAPLOT, AXUM, FIG.P, and many others) is that for certain biochemical
reaction mechanisms the "user specified equation" cannot be derived in
principle.  In the example above, you cannot -in principle- derive an
equation which would give the concentration of P at equilibrium (from
total concentrations and equilibrium constants).  

Clearly, if you cannot derive an equation for the system you are
studying, no data analysis program which asks for a "user specified
equation" can be of any use.  That is also why DYNAFIT was developed for
the chemical, biochemical, and biological community.


_____________________________________________________________
Petr Kuzmic Ph.D. * BioKin Ltd. * Madison, WI 53708-8336, USA
pkuzmic at biokin.com * http://www.biokin.com * 608.256.1269 fax




More information about the Bio-soft mailing list