2. PBPK Modeling
Chemical health risk
assessments
Used to quantify
absorption, distribution,
metabolism, and excretion
(ADME)
Compartments are
specific
Intrinsic and Extrinsic
factors
In relation to chemical
exposure
In silico vs in vitro
Definite use in
pharmacokinetics
3. PBPK Modeling
Models vary based on complexity
Compartment = theoretical value for a chemical
Connections indicate how each parameter calculates
another
2 compartment 6 compartment Several Compartments
4. Parameters
Used to influence organ flows
and partitioning into
compartments: “factors”
related to uptake/circulation
and elimination (or ADME)
Contains descriptors, such as
molecular weight or surface
area( MW or TPSA)
Derived from different
chemical properties
Physiological, Chemical,
Tissue specific
Important!!!!!!!!!! in PBPK
modeling
Examples: Absorption rate
5. Molecular Descriptors to derive
ADME-specific parameters
Chemical descriptors
used to predict
ADME models based
on known value of a
(ADME) response
variable
Chemical structure
and biological
activity
Calculate descriptors
for chemicals in a
database
Using Molecular
Operating
Environment
(MOE)
9. Background/Our Goal
Most experimentation is done on
real life organisms
In silico models are not favored
PBPK modeling doesn’t use real
organisms
Saves lives and money
Create a mobile app that is easily
assessable
Prevents loss of organisms’ lives
10. Methods
Prepare Spreadsheet
Initial Preparation
http://dogwood.rtpnc.ep
a.gov/
Computer Version
Goal: transfer to mobile
app range
12. Methods
Datasets were first identified in the computer
toxicology book, curated, then modeled in MOE
Datasets used: Clearance_Oral, Human Clearance,
Hepatic Clearance, Human Intestinal Absorption,
Human Oral Intestinal Absorption
Descriptors: Hba hydrogen bond acceptor count
(a_acc), Hbd hydrogen bond donor count
(a_don), molecular weight (MW), octanol water
partition coefficient (logP), topological polar surface
area (TPSA), fraction of rotatable bonds
(b_rotR), Number of atoms (a_count)
23. Conversion to a Mobile Device
SpreadsheetConverter
Hid specific sheets
Simplified the spreadsheet to fit into the smaller area
Converted spreadsheets to URL compatible
Created a tiny.url for the newly made webpage
QR code then calculated for the specific URL
End-user of package is now able to view
24. URL and QR Code
http://goo.gl/UDR4U http://goo.gl/3X0pX
ADME by Analog App Physiology App
27. Works Cited
"Assessment of chemicals - Organisation for Economic Co-operation and Development." Organisation for Economic Co-
operation and Development. OECD, n.d. Web. 12 July 2013. <http://www.oecd.org/env/ehs/risk-assessment/intro
ductiontoquantitativestructureactivityrelationships.htm>.
MacDonald, Alex J., and Neil Parrott. "MODELLING AND SIMULATION OF PHARMACOKINETIC AND
PHARMACODYNAMIC SYSTEMS - APPROACHES IN DRUG DISCOVERY." Beilstein-Institut. Beilstein-Institut
Workshop, 22 July 2005. Web. 16 July 2013. <www.beilstein-
institut.de/bozen2004/proceedings/MacDonald/MacDonald.htm>.
U.S. Environmental Protection Agency, Office of Research and Development. (2008). Uncertainty and variability in
physiologically based pharmacokinetic models: Key issues and case studies (EPA/600/R-08/090). Washington, DC:
National Center for Environment Assessment.
Zhao, P. Food and Drug Administration, Center for Drug Evaluation and Research. (2011). Applications of physiologically
based pharmacokinetic (pbpk) modeling and simulation during regulatory review (21191381). Retrieved from Office of
Clinical Pharmacology website: http://www.ncbi.nlm.nih.gov/pubmed/21191381
Notas do Editor
ADME-absorption, distribution, metabolism, excretion; all interlinked processes
compartments-separated into categories like organs, organ systems, absorption paths, etc
Intrinsic: inside and natural; Extrinsic: outside and not apart of the nature of a person
examples: intrinsic-age; extrinsic-drug to drug interactions
In silico: computer experimentation; in vitro: actual experimentation on an organism
We created a needed assessment survey for a variety of needs, such as different types of organisms, organs, and absorption rates. For example, the first bar graph indicates that the most observed organism is the rat. By discovering which organism was most wanted by the modelers, we now knew which parameters should be most important. Additionally, most of the PBPK modelers surveyed did not want in silico models for modeling parameters, an important point for my project.
These are different icons that I created. The general types of icons that were created were different organs, organisms, and absorption rates.
This tab from the spreadsheet is on Weight Estimate
This is taking the chemical and its parameters and estimating the stage of life the human is in and its current weight.
My job was to shrink this table down into one that would fit on a mobile device.
For this, I kept on the human’s data and calculated only human values.
I followed the same process to calculate the different descriptors.
For these normalized values for the parameters in each chemical, I took the maximum and minimum value of each descriptor.
Then, for each descriptor value, I plugged it in an equation, along with the max/min value to calculate the normalized values
This table is used to find the distance between two points with normalized attributes. To do this step, we utilized an equation that is similar to the Pythagorean theorem.
On this slide, we wanted to compare all of the chemicals in this specific dataset to a “test molecule” represented at the left.
This method allowed us to find the top five chemicals that are similar to another chemical in dimensional space.
These histograms graphed all of the descriptor values and the observed values for the chemicals. This helped me calculate the decision tree classifier by demonstrating how all of the data is distributed
This process is how to calculate the decision tree classifier. After calculating descriptors, I created a class named “EF_CLASS”. Using the data found in the histogram, I sorted all of the chemicals into three different bins: 0,1, and 2. After sorting the chemicals into bins, I compute a classification model under the class “EF_CLASS”. I selected all of the descriptors I calculated and created the model. Indicated here, this is the misclassifcation rate. This number indicates that the rate is 16.2% inaccurate, which is also 83.8 percent accurate.
After computing each classifier, I was able to hand draw each classification tree for the different datasets. At this top tree for Clearance_Oral, the separation is based on if the chemical in question followed each equation. If it did, then it branched off to the right, and if it didn’t, it branched off to the left. At the end, the “total” is the misclassification rate. Clearance_Oral’s dataset had a misclassification rate of 18.4%, which also indicates that it is 81.6% accurate. This then indicates that the calculator would be at least 80% accurate each time it computed a value for a desired chemical.
This dataset is known as the Obach dataset and it contains 671 chemicals.
The bottom row represents all the other parameters that need to be calculated for this specific dataset. Each parameter is also calculated for the chemical listed above.