ADME WorkBench is a versatile software application providing flexible, robust pharmacokinetic modeling by integrating state-of-the-art absorption, distribution, metabolism and excretion methods. Designed for research applications in toxicology, pharmacology and biotechnology, ADME WorkBench supports pharmacokinetic prediction from in vitro and/or in vivo data for drugs and environmental chemicals. The ADME WorkBench user interface offers a highly optimized workflow for predictive pharmacokinetics, while allowing ample flexibility to adapt to specific research needs. The pharmacokinetic models used in ADME WorkBench are based on research resulting from the PhRMA CPCDC Initiative on Predictive Models of Human Pharmacokinetics, and described in a series of articles published in the Journal of Pharmaceutical Sciences in 2011. Extension of the models and implementation in ADME WorkBench has resulted from an ongoing scientific collaboration between Dr. Patrick Poulin and AEgis Technologies.
ADME WorkBench assembles a number of state-of-the-art methods for PK prediction. This collection of methods allows a wide variety of data to be used for PK prediction. These methods are configured and data is specified in an intuitive user interface in which compound and PK data is entered using a familiar spreadsheet format. Predictions for compounds may be run in either single- or batch-mode. All models and scripts are delivered with the product, enabling you to inspect and modify the models to suit your particular research needs.
ADME WorkBench Features and Capabilities
If you’re comfortable with a spreadsheet, you’re minutes away from advanced in silico PK prediction.
All compound and in vivo data is entered in a familiar spreadsheet-based user interface, meaning you’ll be using state-of-the-art techniques to generate PK predictions within minutes, even for large sets of compounds. For individual compounds, you can run simulations interactively to find out how changes in parameters of interest affect PK profiles.
Intuitive, Efficient PK Analysis Workflow
The ADME Workbench User Interface (UI) is a Windows Desktop application which conforms to common conventions for popular Microsoft Windows applications, including a main menu, toolbar, status bar, tabbed document area and properties (settings) grid.
PK studies are defined by a list of chemicals, where the chemical-specific properties are entered into a spreadsheet. This spreadsheet is the primary “document” associated with a PK study (also called a “project”) in ADME Workbench. Each row in this spreadsheet corresponds to a compound to be included in the study. Each column corresponds to a chemical-specific property required by the selected prediction methods.
Project-wide settings (e.g., the specific prediction methods to be used) are specified in a project settings grid; the columns in the “compounds” spreadsheet adjust to reflect the compound-specific data required by the selected prediction methods. This combination of a familiar spreadsheet-based user interface and input fields which adapt to the particular prediction methods selected for a specific study enable a highly efficient workflow for predictive PK analysis.
There’s a reason we chose the name “WorkBench.” ADME Workbench employs a variety of predictive PK methods, depending on the data you have available and the predictions you need to make. Allometric techniques based on the Wajima method are available when preclinical in vivo data is available. On the other end of the spectrum, a sophisticated whole-body PBPK model can be used for analyses based on in vitro-in vivo extrapolation.
ADME Workbench supports two general approaches for predicting in vivo pharmacokinetics: allometry based on the Wajima method and Physiologically-Based Pharmacokinetics (PBPK). Each of these methods in turn uses a variety of sub-methods to compute necessary pharmacokinetic quantities based on the available inputs. In general, selection of a prediction method is determined by the nature of data available for input. Allometry requires in vivo blood pr plasma time course data for one or more preclinical species. PBPK models can utilize a wide variety of data, but are most useful when only in vitro and physico-chemical data is available for prediction.
The Wajima method is based on the observation that blood/plasma concentration versus time curves are similar in shape across different species. Moreover, if these curves are properly normalized (i.e., time and concentration axes properly scaled), the curves can be superimposed. In this case, the time axis may be normalized by dividing the time coordinate by the mean residence time (MRT) and the concentration axis by the steady-state concentration (Css). Thus, to predict IV kinetics in humans, the concentration curves for preclinical species may be normalized by using the measured MRT and Css, and then scaled to predict human PK by multiplying the time and concentration coordinates by the predicted human MRT and Css. Human MRT and Css are computed using the relations MRT = Vss/CL and Css = Dose/Vss, meaning a prediction of human Vss and CL (the clearance) are required.
ADME Workbench provides a number of methods for predicting human Vss and CL from preclinical in vivo data.
To determine PK for orally-delivered compounds, values for Ka (first-order absorption constant) and FABS (fraction oral dose absorbed) are estimated from preclinical species data. These values are then averaged to estimate the human Ka and FABS.
Physiologically-based Pharmacokinetics (PBPK)
The PBPK method utilizes a generic, whole-body PBPK model to predict tissue and sub-tissue level concentrations in a variety of tissues. The PBPK model used in ADME Workbench is highly configurable; the sub-models used to predict absorption, distribution and metabolism/elimination can be individually to use a variety of techniques depending on the particular input data available. For example, the absorption module of the ADME Workbench PBPK model can use either a first-order absorption formula, where the rate and fraction absorbed are estimated from preclinical in vivo data, or it can use a multi-compartment GI system model (ACAT) which predicts absorption by modeling the dissolution and permeation of a compound across the intestinal tissue based on the the physico-chemical properties of the compound and Caco-2 permeability information.
Since the PBPK model include representations for individual tissues, predictions can be made of concentration profiles at the tissue level, or at the level of cells or interstitial fluid. By changing the appropriate physiological parameters, the model can be reparameterized to predict kinetics for non-human species.
Source code for all models used in ADME Workbench is delivered and installed with the application, meaning you can examine and customize the model code to suit your particular scientific objectives. In other words, you’re not dependent on us to make modifications to the models if you need to add new tissues or additional metabolic pathways.
The pharmacokinetic models used in ADME Workbench are composed of systems of nonlinear ordinary differential equations (ODEs). To facilitate understanding and modification of these models, they have been developed in a programming language specially designed for constructing mathematical models based on systems of ODEs. This language, called the Continuous Simulation Language (CSL) is based on the CSSL (Continuous Systems Simulation Language) standard, and is implemented in the ACSL and acslX general-purpose simulation tools. These tools translate and compile models specified as systems of equations in the CSL language, and produce executable libraries (Windows DLLs) which are loaded into the ADME Workbench process. By translating the text-based description of the PK models into machine-executable format, the ADME Workbench models run extremely fast. Moreover, the use of the equation-based textual CSL language means end users of ADME Workbench may examine and modify the model code.
A Powerful, Flexible Computational Engine
ADME WorkBench is built on the acslX computational engine, a software environment for modeling, simulation and analysis of complex nonlinear systems and processes. acslX features a rich modeling language and a wide range of ro-bust algorithms for the numerical solution of ordinary differential equations, differential algebraic equations, and discrete event behavior. Models are compiled into executable binary code ensuring the fastest possible performance and support for extremely large or complex models.
As the descendant of the ACSL family of modeling and simulation tools, in use for over 30 years and now widely used for PBPK and PK/PD modeling, acslX supports both textual and graphical languages for specifying complex biological models, and a Matlab-like scripting language for specification of parameter estimation, sensitivity analysis, Monte-Carlo, or Markov-Chain Monte Carlo analyses. For computationally intensive problems, acslX includes support for cluster/grid computing. Data integration capabilities with Microsoft® Excel and Access are provided, and a PBPK/PK/PD toolkit is also available which includes linear and nonlinear PBPK models of various organs, exposure models for inhalation, oral and dermal exposures, and a variety of PD dose response models.
Mathematical and analytical capabilities of acslX include:
- Equation-based modeling of ODEs, DAEs and dhybrid continuous/discrete iscrete systems
- Maximum-likelihood based parameter estimation using a variety of maximization algorithms
- Local and global sensitivity analysis using the extended Fourier Amplitude Sensitivity Test (FAST) and the Morris Screening method
- Bayesian analysis using MCMC sampling, including a language for the specification of hierarchical statistical models
- Monte Carlo analysis for analysis of the variability of model predictions, with support for a variety of statistical distributions
- A flexible scripting language based on the Matlab industry standard, with hundreds of mathematical and analytical function
The PhRMA CPCDC Initiative on Predictive Models of Human Pharmacokinetics recently published a series of papers in the Journal of Pharmaceutical Sciences assessing the predictability of human PK from preclinical data, and comparing various prediction methods. As part of that research effort, a dataset of over 100 compounds was assembled and a variety of empirical and mechanistic modeling approaches were evaluated on their ability to predict human PK properties and clinical data. Specifically, the results of the PhRMA initiative provided guidance on:
- Applicability of general classes of prediction techniques, including whole-body and lumped physiologically-based pharmacokinetic (PBPK) models, and allometry based on the Wajima approach
- Methods for predicting human clearance and steady-state volume of distribution from preclinical data
- Methods for predicting tissue distribution at the macro level (whole organ) and micro level (cell, interstitial fluid) for bound or unbound drug
- Oral absorption models, including the Advanced Compartmental Absorption and Transit (ACAT) model
- How and when to apply the above methods, based on available data and predictive requirements
ADME WorkBench integrates and extends the methods identified in the PhRMA publications to provide a flexible, accurate ADME prediction platform based on industry best practices.
For more information about the PhRMA group, visit http://www.phrma.org.
Citations for the series of articles appearing in the Journal of Pharmaceutical Sciences describing the results of the CPCDC Initiative can be found here.