RevK – Reverse dosimetry and PBPK prediction
This case-study demonstrates and documents the use of a web interface to physiologically-based pharmacokinetic models for forward and reverse dosimetry calculations. Forward calculations compute internal concentrations from given exposure doses. Reverse calculations compute exposure doses from internal concentrations or measured biomarker levels (e.g., urine concentration data). The result of those calculations can be used in risk assessments to help with in vitro to in vivo extrapolations or interspecies extrapolations.
Three tools have been developed for this case-study at NTUA and have been integrated into the OpenRiskNet infrastructure through the Jaqpot web-based computational platform. More specifically, the popular high-throughput toxicokinetic (httk) R package and the PKSim software tool for whole-body physiologically based pharmacokinetic modeling were integrated, but we also developed infrastructure for developing and deploying user-defined model.
For each of these three web tools, simulations are performed and results are presented for reference chemicals or drugs, namely Imazalil for the httk model, Diazepam and Chlorpyrifos for showcasing the In-house R PBPK workflow and Theophylline for the PKSim model. The exposure scenarios chosen are in the range of corresponding environmental or therapeutic levels.
The objective of this case study is to demonstrate and document the capabilities of the OpenRiskNet-developed web-services for Physiologically Based PharmacoKinetic (PBPK) modeling with illustration of both forward and reverse dosimetry predictions.
PBPK models offer a methodology for predicting the internal distribution and exposure of a compound in an organism. Their nature is mechanistic; they consist of compartments representing real organs and tissues, whose number varies based on the target substance, species, administration route and available information. A common approach is to incorporate in the model the main body tissues, i.e. brain, heart, kidney, skin, spleen, liver, lung, gut, bone, adipose and muscle (Jones et al., 2013). Nevertheless, the dimensionality of a PBPK model can be reduced using lumping methods (Pilari et al. 2010; Nestorov et al., 1998). In most cases, PBPK models are utilised for describing the kinetics of a substance in the whole body of a species, thus such models are more formally called “whole body physiologically-based pharmacokinetic” (WBPBPK) models. However, there are models developed to describe in detail the kinetics of a specific organ or body area, which is divided into separate subcompartments. This modeling approach is called “partial” PBPK models (Sturm, 2007).
PBPK models have inherent advantages due to their mechanistic nature. Firstly, they enable predictions of concentration/mass profiles of individual organs and not just plasma. In addition, their relation with physiology and modularity facilitate the integration of literature information, making predictions prior to in vivo experiments possible (Nestorov, 2003). Lastly, their biggest advantage is the ability to perform inter-species (e.g. from rat to human) or intra-species (e.g. from adults to children) extrapolation through scaling methods.
Risk assessment framework
The application frameworks are, for example: REACH risk assessments; SEVESO II directive on safety around industrial plants; Internal chemical, cosmetic, or pharmaceutical company assessments of workers’ safety, or consumer’s safety. All those require integration and extrapolation of in vitro and/or in vivo data on animals to assess human risks.
Databases and tools
We use open source software able to implement PBPK models within the Jaqpot platform (Chomenidis et al., 2017): httk package (Pearce et al., 2017), PKSim (Willmann et al., 2003), and an in-house R client for custom PBPK modelling. The Jaqpot biokinetic services are used to publish the PBPK models as web services. Service clients are developed in the R language. Databases of parameter values are provided by the httk R package, and the PKSim model.
Implementation of the chosen PBPK model as web services:
PBPK models for a specific class of chemicals and animal species can be selected by the user from a particular PBPK modelling environment (e.g., httk in R, PKSim).
The chosen PBPK model is exposed as a web service using the Jaqpot modelling platform. This is possible through the Jaqpot Protocol of Data Interchange (JPDI) which allows to dynamically and seamlessly incorporate practically any algorithmic implementation into Jaqpot. The protocol specifies the form of data exchange between Jaqpot services and third party algorithm web service implementations. The Jaqpot framework already provides wrappers for the R language and the Python language. Integration with R is made possible through the OpenCPU system, which defines an HTTP API for embedded scientific computing based on R, although this approach could easily be generalized to other computational back-ends (Ooms, 2014). OpenCPU acts as a wrapper to R that is readily able to expose R functions as RESTful HTTP resources. The OpenCPU server takes advantage of multi-processing in the Apache2 web server to handle concurrency. This implementation uses forks of the R process to serve concurrent requests immediately with little performance overhead. By doing so it enables access to those functions on simple HTTP calls converting R from a standalone application to a web service.
Demonstration of PBPK models that have been exposed as web services:
The three simulation tools (httk, PKSim and user-specified) are demonstrated with Imazalil, Theophylline, Diazepam and Chlorpyrifos in rainbow trout respectively.
For Imazalil and Theophylline, we start by identifying relevant human exposures (e.g. from ExpoCast, or published literature) to be used in forward dosimetry. For Diazepam and Chlorpyrifos, reverse dosimetry is examined; we identify (e.g. from the US NHANES database, or the scientific literature) typical blood or urine concentrations found in humans to be used as input to the exposure dose reconstruction.
Each model is parameterized using user-specified or pre-programmed tabulated physiological data. For forward dosimetry predictions, each model is run with the given exposure scenario to predict internal concentrations after 24 hours, while for reverse dosimetry, the model is run forward iteratively with user set exposures so as to match the input biomarkers (that is: manually invert the model). The external exposure level leading to data-matching biomarker level is recorded as final estimate.
Several implementations of this case study are described (see details in the case study report linked below):The first implementation uses httk and Imazalil. We describe all the steps required to develop the models as web services through the Jaqpot API or the Jaqpot GUI.
The second is a generic OpenRiskNet framework , which can be used with custom-made PBPK models. Two examples are provided, a PBPK model for Diazepam in humans, and a generic (i.e. not substance-specific) PBPK model in fish. In the case of diazepam, the tools were used to analyse biomonitoring data regarding diazepam blood levels in drivers. In the case of the fish PBTK model, exposure levels which lead to in vitro effects on biomarkers in liver were estimated.
The last implementation is the integration of a PBPK model for Theophylline, originally developed in the PKSim software. We describe all the steps required to develop the model as a web service through the Jaqpot API.
We are providing all the steps required to perform dosimetry through the Jaqpot GUI using the custom-made model as examples.
The results of this case-study demonstrate that the OpenRiskNet framework can be used as a central e-platform for the biokinetics community, where the users can publish, share, search and use PBPK models.
Currently available services:
Generate, store and share predictive statistical and machine learning modelsService type: Analysis tool, Processing tool, Trained model, Model generation tool, Model, Data mining tool, Service
Service type: Database / data source
Predict ADME/PK with ConfidenceService type: Application, Software, Service
Interactive computing and workflows sharingService type: Visualisation tool, Helper tool, Software, Analysis tool, Processing tool, Workflow tool
Computation research made simple and reproducibleService type: Database / data source, Visualisation tool, Software, Analysis tool, Service, Workflow tool
This case-study demonstrates and documents the use of a web interface to physiologically-based pharmacokinetic models for forward and reverse dosimetry calculations. Forward calculations compute internal concentrations from given exposure doses. Reverse calculations compute exposure doses from internal concentrations or measured biomarker levels (e.g., urine concentration data). The result of those calculations can be used in risk assessments to help with in vitro to in vivo extrapolations or interspecies extrapolations. Three tools have been developed for this case-study at NTUA and have been integrated into the OpenRiskNet infrastructure through the Jaqpot web-based computational platform. More specifically, the popular high-throughput toxicokinetic (httk) R package and the PKSim software tool for whole-body physiologically based pharmacokinetic modeling were integrated, but we also developed infrastructure for developing and deploying user-defined model. For each of these three web tools, simulations are performed and results are presented for reference chemicals or drugs, namely Imazalil for the httk model, Diazepam and Chlorpyrifos for showcasing the In-house R PBPK workflow and Theophylline for the PKSim model. The exposure scenarios chosen are in the range of corresponding environmental or therapeutic levels.
The OpenRiskNet project (https://openrisknet.org/) is funded by the H2020-EINFRA-22-2016 Programme and its main objective is the development of an open e-infrastructure providing data and software resources and services to a variety of industries requiring risk assessment (e.g. chemicals, cosmetic ingredients, pharma or nanotechnologies). The concept of case studies was followed in order to test and evaluate proposed solutions and is described in https://openrisknet.org/e-infrastructure/development/case-studies/. Two case studies, namely ModelRX and RevK, focus on modelling within risk assessment. The ModelRX – Modelling for Prediction or Read Across case study provides computational methods for predictive modelling and support of existing data suitability assessment. It supports final risk assessment by providing calculations of theoretical descriptors, gap filling of incomplete datasets. computational modelling (QSAR) and predictions of adverse effects. Services are offered through Jaqpot (UI/API), JGU WEKA (API), Lazar (UI) and Jupyter & Squonk Notebooks. In the RevK – Reverse dosimetry and PBPK prediction case study, physiologically based pharmacokinetic (PBPK) models are made accessible for the purpose of risk assessment-relevant scenarios. The PKSim software, the httk R package and custom-made PBPK models have been integrated. RevK offers services through Jaqpot (UI/API).
Nano-QSAR to predict cytotoxicity of metal and metal oxide nanoparticles
Lazar Toxicity Predictions
The aim of this study is to benchmark two Bayesian software tools, namely Stan and GNU MCSim, that use different Markov chain Monte Carlo (MCMC) methods for the estimation of physiologically based pharmacokinetic (PBPK) model parameters. The software tools were applied and compared on the problem of updating the parameters of a Diazepam PBPK model, using time-concentration human data. Both tools produced very good fits at the individual and population levels, despite the fact that GNU MCSim is not able to consider multivariate distributions. Stan outperformed GNU MCSim in sampling efficiency, due to its almost uncorrelated sampling. However, GNU MCSim exhibited much faster convergence and performed better in terms of effective samples produced per unit of time.
Aim: understanding the use, form, inputs and outputs of physiologically based (PBPK) pharmacokinetic models. Presentation of software applications for developing PBPK models. Customising PBPK to individual time-drug concentration data. Creating optimal drug dosage regimens