Early stage structure driven PBPK modelling Daniel Mucs Scientific Computing Group

Early stage structure driven PBPK modelling
Daniel Mucs
Scientific Computing Group
Cyprotex Discovery Ltd., Macclesfield, UK and Boston MA
Presentation Outline
Introduction to Physiologically Based Pharmacokinetic (PBPK) modelling
Cloe® Gateway
Simple PBPK approach
Pattern recognition methods
Summary
Presentation Outline
Introduction to Physiologically Based Pharmacokinetic (PBPK) modelling
Cloe® Gateway
Simple PBPK approach
Pattern recognition methods
Summary
Introduction to PBPK modelling
PBPK models are mathematical simulation models
They are devised to predict/mimic the fate(s) of compound(s) in the bodies of
humans, preclinical species and/or other organisms, i.e. ADME IVIVE
They are expressed as a system of differential equations that are solved
simultaneously
Their primary output is the change over time following dosing of relevant
quantities. e.g.:
The concentration of a compound in the plasma and other tissues.
The amount of a compound eliminated in the urine or the bile
The amount of a compound absorbed from the GI tract lumen.
Introduction to PBPK modelling
PBPK models are being recommended and used increasingly by regulatory
agencies and consortia
FDA Guidance for Industry 2012
ITC Review 2010
A conceptual physiological model used to predict somatic
distribution and elimination
Venous Blood
Infusion
Elimination
Oral dose
Intracellular space
Interstitial fluid
Capillary bed
Gonads
Liver
Kidney
Brain
Fat
Muscle
Heart
Lung
Stomach
Intestines
Arterial Blood
PBPK models combine physiological and compounddependent properties
Compounddependent
PhysChem
properties
Speciesdependent
ADME properties
Measured OR predicted
Physiological
properties
PBPK Models Predict Plasma Concentration
From the predicted plasma profiles, PK parameters can be calculated using standard
algorithms, and compared to the values observed in vivo.
PBPK models can predict summary PK parameters
Cmax observed
Cmax predicted
Presentation Outline
Introduction to Physiologically Based Pharmacokinetic (PBPK) modelling
Cloe® Gateway
Simple PBPK approach
Pattern recognition methods
Summary
Cloe® Gateway
www.cloegateway.com
PBPK Model Inputs for Screening in Early Drug Discovery
Input Property
How Used in Model
Hepatic microsomal intrinsic clearance
(species-dependent)
Scaled to generate a total hepatic
metabolic clearance
Fraction unbound in plasma (speciesdependent)
Used directly.
Blood:plasma ratio (speciesdependent)
Used directly
pKa(s)
Used to calculate charge state in the
plasma, in cells, and in stomach and gut
lumen.
logP octanol/water
Used (with charge) to calculate
intracellular partitioning and
permeation.
Caco-2 permeability
Used to determine permeation across
GI tract wall.
Solubility (buffered)
Used to determine dissolution within
the GI tract
*Cloe® PK V2.1
PBPK* model is superior to well-stirred model for
predicting in vivo clearance in human
Cloe® PK
well-stirred
*Cloe® PK V2.1.6
Model
rs
MFE
Cloe® PK
0.75
2.0
well-stirred
0.52
3.5
PBPK* model is superior to well-stirred model for
predicting in vivo clearance in rat
Cloe® PK
well-stirred
*Cloe® PK V2.1.6
Model
rs
MFE
Cloe® PK
0.63
2.2
well-stirred
0.47
4.1
Presentation Outline
Introduction to Physiologically Based Pharmacokinetic (PBPK) modelling
Cloe® Gateway
Simple PBPK approach
Pattern recognition methods
Summary
PBPK modelling: a simpler approach
Simpler approach
Current approach
Input Property
Input Property
Hepatic microsomal intrinsic
clearance (species-dependent)
Hepatic microsomal intrinsic
clearance (species-dependent)
Fraction unbound in plasma
(species-dependent)
Fraction unbound in plasma
(species-dependent)
Simple physicochemical properties
Blood:plasma ratio (speciesdependent)
pKa(s)
logP octanol/water
Caco-2 permeability
Solubility (buffered)
PBPK modelling: a simpler approach
Advantages
Structurally globally identifiable
Increased speed of execution (by up to 100 fold)
• Optimise on new sets relatively rapidly
• Error analysis and sensitivity analysis faster
• Screen a larger number of compounds aiming for virtual screening
Comparable oral dose Cmax and AUC predictions with less resources, for specific
areas of chemistry
PBPK modelling: a simpler approach
Aims
Create different versions based on available in vitro data
Ultimately a fully in silico approach which does not require any in vitro data
Not needing any in vitro data means even the compound synthesis is non-essential,
so PBPK approach can be used at the lead generation stage
This way PK parameters can be directly related to structural differences
Presentation Outline
Introduction to Physiologically Based Pharmacokinetic (PBPK) modelling
Cloe® Gateway
Simple PBPK approach
Pattern recognition methods
Summary
Pattern recognition methods
Importance of patter recognition in PBPK modelling
Predicting model parameters which are:
not measurable for various reasons
no or minimal data available
Explore and analyse large descriptor (calculated or measured) sets exhaustively
Useful in direct IVIVE predictions as well
Pattern recognition example
Fathead Minnow Aquatic Toxicity Dataset
Compiled by He and Jurs in 2005
Comprises 322 compounds that were experimentally assessed for toxicity to fathead minnow
The dataset contains descriptors and fingerprints from ten different sources:
AtomPair
Daylight_FP
Dragon
Lcalc
Moe2D,
moe2D_FP
moe3D
PipelinePilot_FP
QuickProp
Fathead Minnow Aquatic Toxicity Dataset
Compare models using different
descriptor sets and model building
methods
moe2D gives the lowest RMSE so use
that for final model building
Find optimal number of descriptors
Do n number of experiments with
cross validation
Fathead Minnow Aquatic Toxicity Dataset
Good correlation with observed data
Minimal number of outliers
Initial clustering of the dataset with
this approach is not necessary
Presentation Outline
Introduction to Physiologically Based Pharmacokinetic (PBPK) modelling
Cloe® Gateway
Simple PBPK approach
Pattern recognition methods
Summary
Summary
Physiologically Based Pharmacokinetic (PBPK) models are versatile simulation
models for the prediction of xenobiotic concentrations in plasma and other
tissues following administration by various routes.
Cloe® PK is a generic PBPK model, requiring simple ADME and PhysChem data as
input, and is suitable for use in prioritising compounds during drug development
The simpler model approach could be used in early drug discovery as it requires
minimal in vitro data
The aim is to develop a PBPK model which would only need calculated properties,
so even synthesis is not essential for predicting PK parameters