Dear Steve,
In response to your question on the BB of Thurs., April 26, there are indeed
programs that do the calculations that I think your friends are desiring. Speci
fically, there are programs that will do nonlinear least squares analysis of sys
tems of ordinary differential and algebraic equations that define any kinetic mo
del. There are also many programs facilitating the simulation of the time cours
e of any reaction so defined. The later programs do not normally contain a powe
rful optimizer that efficiently finds the best fit parameters, although a few of
the simulation programs will permit this to be done in a crude fashion. If sim
ulation rather than best fit parameters are what you want, however, the simulati
on programs are much easier to learn and use.
Our CRICF program [Comput. Biomed. Comput. vol. 5, 515-534 (1972) & ibid, vol
. 9, 203(1976)] was the first digital computer program to accomplish the genera
least squares fitting of model equations to reaction rate data. However, it is
quite old now and I would recommend the programs by Carl Frieden [Biochem. J.,
vol. 258, 381-387 & Anal. Biochem. vol. 130, 134-145 (1983)] currently if you ha
ve a VAX/VMS system. The one remaining advantage our program has is its virtuel
y complete portability among types of computers including a version that runs qu
ite nicely on microcomputers (the latter doesn't include the stiff differential
equation integrator and does have some machine dependent I/O). I should also ad
d that we made a mistake in testing Gear's method with stiff differential equati
ons. Therefore, in contrast to our original conclusions, Gear's method is as go
od as or superior to the Treanor method of integrating stiff differential equati
ons that we use. However, we have not yet had rate equations where the differen
ce between these two methods was significant. For non-stiff differential equati
ons, we provide a variable step size, fourth order Kutta method with an error es
timate that works well.
Carl Frieden's two programs are simulation and fitting programs, resp. Howev
er, there are a large number of simulation packages available for both large and
modest size model complexities and large or microcomputers. The National Biome
dical Simulation Resource (NBSR) provides both types - a version for the PC and
access to their host computers providing a range of powerful hardware, including
array processors. Address to the NBSR is Box 3709 Duke University Medical Cent
er, Durham, NC 27710, ph. 800-334-2083 in the USA. Other simulation packages su
itable for microcomputers are available (see commercial software reviews), but I
prefer the NBSR programs.
Best wishes,
H. Olin Spivey, Biochemistry Dept., Oklahoma State University, Stillwater, OK
74074, ph. 744-6192, Bitnet address: BIOCHOS at OSUCC.
