Scope of Modelingan Overview Repository

Meyer Katzper

September, 2006

A Presentation Designed to Provide Awareness of the Variety of Modeling Approaches Available

Goal:

Expand Horizons Regarding Modeling Possibilities

diagram from http://nsr.bioeng.washington.edu/PLN/Members/butterw/JSIMDOC1.6/Admin_History.stx/JSim_Science.stx

Review Activities with Current

Major Modeling Components

Medical

Statistical X

Pharmacokinetic X

Chemical

Toxicological

Research Disciplines Focusing on ModelingDiscipline DivisionMedical

Statistical Quantitative Methods Research Staff

Pharmacokinetic Pharmacometrics Staff

Chemical

Toxicological Informatics and Computational Safety Analysis Staff

Microbiological

Genomic ProteomicBiochemical

MechanismCellular Tissue/Organ Organism

Geneexpression

Proteinsynthesis

ProteinModification

functionalActivity

Adaptation,repair

or damage

Adaptation,repair

or damage

Dysfxn

D

TOXICITY!

Levels and Factors Involvedin Determining Toxicity

Toxicology Data Continuum Representation by Daniel A. Casciano, Ph.D.National Center for Toxicological Research

A Futuristic View of A Reviewers Job

Input the drug entity into the CDER Virtual Person

Obtain

The distribution of the drug in all body organs, with target organs highlighted.

Probable efficacy for the concentration at the target for a given disease status and genomic profile.

Possible adverse events.

Comparative efficacy and safety profile versus competing drugs.

Reviewer Task Verify the computer results. Make a decision on approval and determine labeling.

This presentation will touch on

The World of Modeling. The Multiplicity of Modeling Techniques.

Excerpts on overall approach and techniques.

The Practical - Link to FDA interests.

Computer models can be viewed in different ways and classified according to several criteria

A Major Classification set is

Stochastic or deterministic Continuous or discrete

Lumped parameter or distributed Mathematical or Agent Based

Note chaotic is a special case of deterministic.

CDER has two main modeling groups

Statistical Modelers Pharmacometric Modelers

These are a small subset of the modeling approaches available

Model types Cellular automata

Agent-based simulation

System Dynamics

Micro-Simulation

Neural networks

Genetic Algorithms

Linear Programming

Integer Programming

Dynamic Programming

Combinatorial Optimization

Stochastic Processes

Markov Chains

Finite element

Queuing

A completely different approach arises from researches in network theory.Reference Article --NETWORK BIOLOGY:UNDERSTANDING THE CELL’S FUNCTIONAL ORGANIZATIONAlbert-László Barabási & Zoltán N. Oltvai

Reference for picture (next slide)Jeong, H., Mason, S. P., Barabási, A.-L. & Oltvai, Z. N.Lethality and centrality in protein networks. Nature 411, 41–42

Article athttp://www.nd.edu/~networks/Publication%20Categories/03%20Journal%20Articles/Biology/NetworkBio_Nature%20Rev%20Genetics%205,%20101-113%20(2004).pdf

Figure;| Yeast protein interaction network. A map of protein–protein interactions in Saccharomyces cerevisiae, which is based on early yeast two-hybrid measurements. illustrates that a few highly connected nodes (which are also known as hubs) hold the network together. The largest cluster, which contains ~78% of all proteins, is shown. The colour of a node indicates the phenotypic effect of removing the corresponding protein (red = lethal, green = non-lethal, orange = slow growth, yellow = unknown).Reproduced from REF. 18 ©

 

Tool Name Function

COR Cardiac Electrophysiology Simulator

iCell Cardiac Electrophysiology Simulator

LabHeart Cardiac Electrophysiology Simulator

CMGUI Visualisation

CMISS PDE Simulator

ViPEr Visualisation

Continuity PDE Simulator

Gepasi Biochemical Simulator

Systems Biology Workshop A framework to facilitate application intercommunication

Virtual Cell Biological Simulator

BioPSE Electrical Field Simulator, and Visualisation

CESE Cardiac Electrophysiology Simulation environment

List of software from http://www.physiome.org.nz/tools

Alternate Techniques

The modeling technique and approach to deal with questions which appear quite similar may be radically different. Consider substance toxicity which EPA and FDA must deal with.

The EPA in carrying out its toxicity studies makes use of Physiologically Based Pharmacokinetics. (PBPK)

Their Scientific Advisory Panels deal with PBPK/PD models for the evaluation of pesticides.

CDER does not use PBPK models.

EPA is currently advocating mode of action (MOA) models in its guidelines. Its use is in carcinogenic hazard identification. This methodology is still being elucidated.

See http://birenheide.com/sra/2005AM/program/singlesession.php3?sessid=W14

SamplePhysiologically Based Pharmacokinetic Model

reference http://www.biomedcentral.com/1472-6904/2/5

SamplePhysiologically Based Pharmacokinetic Model

Reference http://www.biomedcentral.com/1472-6904/6/1

The various modeling approaches each have distinct world views and representations. Adjusting to a different view is difficult.1

Method I Mainly Use: Systems Dynamics2

Software: STELLA©, a graphically oriented modeling tool used for simulation. Also MADONNA©

Use Requirements: Specify the main elements of concern and their interactions.

Vague Definition: System dynamics is an approach to modeling the dynamics of entities, which usually interact strongly with each other.

Note – Many problems can be programmed in EXCEL.

1 “Different categories of programming languages elicit quite different modes of thinking and problem solving.”2”I write programs in Lisp for the same reason I write prose in English – not because it’s the best language, but because it’s the language I know best.”Quotes from B. Hayes computer science column in American Scientist, July-August 2006.

STELLA is a VISUALLY ORIENTED MODELING ENVIRONMENTfor

DIFFERENCE AND DIFFERENTIAL EQUATIONS.

Laying Out And Connecting Icons In Stella as shown

Is Equivalent To Writing Differential Equations

dY/dt = -k*Y

Multiple linked differential equations with feedback can be used to create a Systems Dynamic model. The model can be solved numerically.

Y

k

Flexibility of System Dynamics – Variety of Possible Applications

Models posted by James L. Hargrove, Ph.D.

Colon Cancer

Osteoporosis

Energy Balance

Cholesterol Metabolism

Gene Expression

A Text by James L. Hargrove Dynamic Modeling in the Health Sciences

provides many more examples. http://www.arches.uga.edu/~jhargrov/

                                                  

Pain and Pills ExampleOne of the observed phenomena with chronic diseases is dose creep.There are a number of possibilities as to why this occurs.Results from a model of chronic diseases with dose creep is shown below.In this model pain is assumed to get worse as the disease progresses.To obtain the same degree of pain relief the subject increases the dose.

4:30 AM 4/18/1940

0.00 5.00 10.00 15.00 20.00

Time

1:

1:

1:

2:

2:

2:

0.00

0.50

1.00

0.00

2.50

5.00

1: Progressive Disease 2: Analgesic dose

1

1

1

1

2

2

2

2

Progression

REMODELING

BONE

ENDOCRINESYSTEM

INTRA-CELLULARCALCIUM

EXTERNALINTAKE

Drug Interactionwith Environment

HealthStatus

ConceptualModel:EvaluatingA Drug inContext

An excellent introductory article which I highly recommend is Mathematical Modeling in Surgical Research. It is available at http://malthus.micro.med.umich.edu/lab/pubs/buchman.pdf

The article discusses a variety of modeling platforms. However,it features a STELLA example.

Before I knew about this article, I independently worked on modeling creatinine as I considered it important for the study of drug dosage during pediatric development.

The STELLA explanatory example given in this article is of creatinine. Compare it with what I have shown you.

Relevant Academic Modeling

The first computational model ofnicotine addiction is described in theJanuary 24, 2006, issue of the Proceedingsof the National Academy of Sciences.“We wanted to tease out differentparts of addiction,” says lead authorBoris Gutkin, PhD, a research scientistat the Pasteur Institute in Paris,France. “The goal is to put behavioraland neural information together and tosee how the behavioral effects are producedby the neural effects.”

Source

http://www.biomedicalcomputationreview.org/

http://biomedicalcomputationreview.org/2/2/index.html

Commercial Firms and Model Development

http://www.rosapharma.com/consulting/clinical.htm

Quote from web site.

“Drug-Disease Modeling (DDM) and Clinical Trial Simulation (CATD) are proven ways to understand drug action and optimize clinical trial design.

We incorporate physiologic and mechanistic knowledge with proven statistical approaches such as dose-response or PK/PD modeling. With this ability Rosa makes it possible for its clients to confidently extend their understanding and predictive capabilities beyond existing clinical data sets and gain deeper scientific understanding of variability and uncertainty.”

Drug-Induced Liver Injury (DILI) PhysioLab

“The US Food and Drug Administration (FDA) and Entelos are engaged in a collaborative effort to bring the benefits of Entelos PhysioLab technology and research expertise to the field of drug-induced liver injury (DILI).”

Summary of review Process Goals

Enhance Sound Science

Improve understanding of efficacy

Improve understanding of risk

Innovate to improve the drug approval process

Identify and deal with emerging risk issues

Modeling can assist with all these goals

Final message

There are many different modeling approaches.

You can not try an approach you are unacquainted with.

It is desirable to know about meaningful options.

It is desirable to have modeling activities tracked.

NOTE – Expanded information on various approaches is provided in separate presentations.