Back to the Top
My name is Dr. Shlomo Almog and I am the director of the Clinical
Toxicology and Pharmacology Institute at H.Sheba Medical Center,
I wonder if you could take a few minutes to advise me about the following
problem, which often arises when I need to fit pharmacokinetic models to
data for individuals. I use the MK MODEL package (Biosoft , England,
written by N. Holford, version 4.05 for dos) to explore complex models such
as extravascular multiple dozing (parent drugs and metabolites).
Oftentimes (after a very long and frustrating wait) I get amessage that
my initial parameter estimates were unsatisfactory, or that
convergence was not achieved, or that division by zero was attempted,
or that the model did not match.
My question is, do you know of a software package (maybe a new version of
MKMODEL) which could save me some of these headaches? I'm thinking of
something with perhaps a sophisticated way of computing initial parameter
estimates, or with the capability to choose among competing models earlier
on in the iterative process, or (possibly) alternative numerical algorithms
to those in MK MODEL. The package doesn't need to be couched in
pharmacokinetic terms; any software dealing with multicompartment models
would be appropriate.
Thanks in advance for your kind help.
Dr. Shlomo Almog
Back to the Top
Reply to Shlomo Almog concerning problems fitting PK models-
Your problem sounds more fundamental than a choice of software package.
The messages you describe are commonplace for many fitting algorithms
and reflect the numerical difficulties of fitting many parameters
simultaneously to multiple data sets.
It is likely you would get the same messages from other packages unless
your initial parameter values are in the correct range. It might be
useful to first simulate your model with your current initial
parameters, and then look to see if your solution is even (as we
Americans say) in the right ballpark, that is, to see if you are
reasonably close. If not, you can use your own knowledge of the system
to assign some improved initial estimates, then simulate again.
Once you are close, you can invoke the least-squares fitting procedure
and very likely achieve convergence.
If you want to try another modeling package to compare to MK=DDModel, you
might look at SAAM 31 or SAAM=DDII or ACSL BioMed. URLs for all of these
are on the Useful Links page on my website:
Robert D. Phair, Ph.D. rphair.aaa.ix.netcom.com
BioInformatics Services http://www.webcom.com/rphair
Partnering and Outsourcing for Computational Biology
Back to the Top
Have you tried KINETICA, a new software program for pharmackinetics ?
For full information contact Philippe Michon at
pmichon.aaa.bioscience.com. Or call directly +126.96.36.199.13.25.
You can also get further information on this and other products via
the website at www.bioscience.com. You can download a full demo
copy of the Windows 95/NT version.
Good luck !
You can contact us by Email:
Web Site: http://www.bioscience.com
FTP Site: ftp://ftp.bioscience.com
(db - the Kinetica page is at
http://www.bioscience.com/html/software/kinetica.htm and now listed on
PharmPK Discussion List Archive Index page
Copyright 1995-2010 David W. A. Bourne (firstname.lastname@example.org)