Back to the Top
Dear all,
I am looking for a study devoted to the use of modeling neural
networks in pharmacokinetics. I will greatly appreciate receiving any
advice.
With best regards
Maria Durisova
www.uef.sav.sk/durisova.htm
Back to the Top
Dear Maria Durisova,
You can search this topic on pubmed. In my memory, I read a
review published on JPKPD in about 1991. I am not sure.
Guangli Ma
Back to the Top
The following message was posted to: PharmPK
Maria
You'll find extensive work published in the last ten years by Peter
Veng-Pedersen, Michael Brier, Ajaz Hussain and Joga Gobburu (among
others) on ANN's applied to PKPD.
Best
Luis
--
Luis M. Pereira, Ph.D.
Assistant Professor, Biopharmaceutics and Pharmacokinetics
Massachusetts College of Pharmacy and Health Sciences
179 Longwood Ave, Boston, MA 02115
Phone: (617) 732-2905
Fax: (617) 732-2228
Luis.Pereira.-at-.bos.mcphs.edu
Back to the Top
Dear Maria:
When you consider various methods of analyzing PK data, such
as neural networks, parametric population modeling, nonparametric pop
modeling, etc., it is useful to ask -
1. Is the method statistically consistent? that is, if I study
more patients, will my parameter estimates, including the variances
and covariances, get closer to the true values?
2. How rapid is the stochastic convergence of the method? Does
it only need 4 times the original number of subjects to reduce the
estimates of the parameter SD's by half, as theory says, or are more
subjects really needed?
3. What is the statistical efficiency of the method, or its
inverse, the relative precision (SD, for example) of the parameter
estimates compared to the true values?
There are now parametric population modeling methods that
obtain exact likelihood estimates. Nonparametric methods such as PEM
and NPAG also do this. Methods obtaining exact likelihoods generally
are consistent, have stochastic convergence matching theory, and have
good statistical efficiency and precise parameter estimates, as shown
in the table below, taken from reference #1 below.
Estimator Relative efficiency
Relative error
DIRECT OBSERVATION 100.0 % 1.00
PEM 75.4% 1.33
NPAG 61.4% 1.63
NONMEM FOCE 29.0% 3.45
IT2B FOCE 25.3% 3.95
NONMEM FO 0.9% 111.11
Table III. Results of First Monte Carlo Simulation. Comparison of
the relative statistical efficiency and relative error in parameter
estimation for the mean value of V, the apparent volume of
distribution, of the PEM, NPAG, NONMEM FOCE, IT2B FOCE, and NONMEM FO
population Modeling Methods.
Methods using approximations of the likelihood, such as the
iterative 2 stage Bayesian, and NONMEM, both of which use the FOCE
approximation, (and NONMEM also uses the much worse FO
approximation), do not have these properties. It is useful to ask if
neural nets have these desirable properties or not.
Further, is your task ended when you describe and report
your results, or will your results then be used for patient care in
planning, monitoring, and adjusting dosage regimens of various
potentially toxic drugs? In that case, you might wish to use multiple
model dosage design, which can estimate in advance the weighted
squared error with which a regimen will hit the target goal, and find
the regimen which specifically minimizes that error, thus hitting
your target with maximum precision.
I don't know about the properties of neural nets with regard
to the above considerations. However, some relevant references you
might examine from our USC Laboratory of Applied Pharmacokinetics are -
1. Bustad A, Terziivanov D, Leary R, Port R, Schumitzky A, and
Jelliffe R: Parametric and Nonparametric Population Methods: Their
Comparative Performance in Analysing a Clinical Data Set and Two
Monte Carlo Simulation Studies. Clin. Pharmacokinet., 45: 365-383,
2006. One table from that paper is shown below.
2. Jelliffe R, Schumitzky A, Bayard D, Milman M, Van Guilder M,
Wang X, Jiang F, Barbaut X, and Maire P: Model-Based, Goal-Oriented,
Individualized Drug Therapy: Linkage of Population Modeling, New
"Multiple Model" Dosage Design, Bayesian Feedback, and Individualized
Target Goals. Clin. Pharmacokinet. 34: 57-77, 1998.
3. Jelliffe R, Bayard D, Milman M, Van Guilder M, and Schumitzky
A: Achieving Target Goals most Precisely using Nonparametric
Compartmental Models and "Multiple Model" Design of Dosage Regimens.
Therap. Drug Monit. 22: 346-353, 2000.
Very best regards,
Roger Jelliffe
Roger W. Jelliffe, M.D. Professor of Medicine,
Division of Geriatric Medicine,
Laboratory of Applied Pharmacokinetics,
USC Keck School of Medicine
2250 Alcazar St, Los Angeles CA 90033, USA
Phone (323)442-1300, fax (323)442-1302, email= jelliffe.-at-.usc.edu
Our web site= http://www.lapk.org
Back to the Top
The following message was posted to: PharmPK
Dear Maria,
Was your question regarding the prediction of pharmacokinetic parameters
from molecular structure, or the fitting or pharmacokinetic data? I
(and I think Luis, based on the references he provided) assumed it was the
former, but I can't find the original post (even searching the archives), and
I see that Roger assumed it was the latter.
Walt
Walt Woltosz
Chairman & CEO
Simulations Plus, Inc. (AMEX: SLP)
42505 10th Street West
Lancaster, CA 93534-7059
U.S.A.
http://www.simulations-plus.com
Phone: (661) 723-7723
FAX: (661) 723-5524
E-mail: walt.at.simulations-plus.com
Want to post a follow-up message on this topic?
If this link does not work with your browser send a follow-up message to PharmPK@boomer.org with "Neural networks in pharmacokinetics" as the subject | Support PharmPK by using the |
Copyright 1995-2011 David W. A. Bourne (david@boomer.org)