The DILIsym® modeling software is designed to be used during drug development to provide an enhanced understanding of the DILI hazard posed by individual molecules and to provide deeper insight into the mechanisms responsible for observed DILI responses at various stages of the development process.
DILIsym® is a “middle-out”, multi‑scale representation of drug-induced liver injury. It includes key liver cell populations (e.g., hepatocytes, Kupffer cells), intracellular biochemical systems (e.g., mitochondrial dysfunction), and whole body dynamics (e.g., drug distribution and metabolism). The model represents physiological data for mice, rats, dogs, and humans. Through the generation of SimPops™ (alternate parameterizations of the model with distribution constraints), the DILIsym® software also includes inter‑individual variability.
By integrating preclinical data, DILIsym® allows you to predict hepatotoxicity for:
- Small preclinical species (i.e., in vitro to in vivo extrapolation, IVIVE, for mice or rats)
- Large preclinical species (i.e., in vitro and rodent data to dog predictions)
- First-in-human clinical trials
DILIsym® is built in the MATLAB computing platform (The MathWorks, Natick, MA). Users may work directly with the MATLAB code or may use a graphical user interface (GUI). The GUI permits the specification of new experiments, including compound, dose, dosing frequency, duration, mechanism(s), and species. Simulations can be performed on individuals or populations. Simulation results can be visualized within MATLAB or can be exported for further analysis using other software.
DILIsym® prospectively supports key management decisions by providing information on experimental or clinical designs that would expose liver toxicity, as well as mechanistic rationale for the underlying biochemical events that would cause liver toxicity. The model retrospectively supports key management decisions by mechanistically interpreting your data and to distinguish innocuous vs. dangerous liver signals.
The current version of the modeling software is DILIsym® v5A, which was made available to all DILI-sim Initiative industry members as of July 2016.
Mitochondrial toxicity modeling
The representation of mitochondrial biology within the model scope includes mitochondrial respiration, proton-motive force, and ATP production in addition to compounds that disrupt these processes. Exemplar compounds will be simulated to refine the model and ensure that the modeled representation of mitochondrial toxicity is consistent with available data. The exemplar compounds include buprenorphine and etomoxir. The model also incorporates associated pathways that participate in the overall hepatic response to mitochondria toxicity in vivo (e.g., glycolysis). Further, a distinct model of in vitro mitochondrial respiration and ATP production (MITOsym®) is represented and provides direct comparisons with in vitro mitochondrial assays.
Innate immunity modeling
The representation of the innate immune response within the model scope includes macrophage (Kupffer) and liver sinusoidal endothelial cell (LSEC) populations. Macrophages and immunomodulatory molecules produced by macrophages have been shown to modify the course of APAP hepatotoxicity in animal models and are also present in human APAP overdose. LSECs regulate immune cell recruitment to the liver and produce molecules that are involved in the regeneration phase of hepatotoxicity. APAP serves as the primary exemplar compound as most of the available data characterizes the role of these cells and molecules in APAP hepatotoxicity. These include liposomal clodronate (for macrophage depletion), a HMGB1 antagonist, a TNF-α antagonist, and exogenous HGF.
Bile acid modeling
The representation of bile acid biology within the model scope includes the dynamics of specific bile acids implicated in hepatotoxicity. The model incorporates bile acid synthesis, uptake, recirculation, and efflux, including key transporters (e.g., bile salt efflux pump or BSEP). Drug-induced changes to bile acid dynamics can increase intracellular bile acids with subsequent hepatotoxicity. Exemplar compounds will be simulated to refine the model and ensure that the modeled representation of bile acid toxicity is consistent with available data. For this purpose, Pfizer has generously made available data on CP‑724714, a HER2 tyrosine kinase inhibitor whose hepatotoxic side effects have been attributed to BSEP inhibition. Other compounds of interest are also being simulated.
DILIsym® modeling software functions include:
- General population samples: The DILIsym® software can be used to test compounds in simulated populations, termed SimPops™. Simulated individuals within the SimPops™ express a wide‑range of response to the various exemplar compounds and can be characterized by extensive variability in their underlying biochemistry.
- DILIsym® humans: A middle-out multi-scale representation of human physiology for assessing potential DILI hazard in patients. Compound pharmacokinetic (PK) and pharmacodynamics (PD) information can be integrated to predict time profiles of liver enzymes (i.e., alanine aminotransferase, aspartate aminotransferase) and other clinical variables (e.g., bilirubin, prothrombin time, INR), as well as tissue properties (e.g., liver mass, GSH content). Alternate hypotheses regarding the downstream mechanisms of drug action can be investigated, including increased reactive oxygen/nitrogen species, ATP utilization, direct hepatocyte necrosis, and inhibition of bile acid transporters.
- DILIsym® dogs: A middle-out multi-scale representation of Beagle dog physiology for assessing potential DILI hazard in dogs. Data from Beagle dogs were prioritized among the available datasets to maximize consistency in the representation.
- DILIsym® rats: A middle-out multi-scale representation of Sprague-Dawley rat physiology for assessing potential DILI hazard in rats. Data from Sprague-Dawley rats were prioritized among the available datasets to maximize consistency in the representation. Compound pharmacokinetic and pharmacodynamics information can be integrated to predict time profiles as described above.
- DILIsym® mice: A middle-out multi-scale representation of C57Bl/6 mouse physiology for assessing potential DILI hazard in mice. Data from C57Bl/6 mice were prioritized among the available datasets to maximize consistency in the representation. Compound pharmacokinetic and pharmacodynamics information can be integrated to predict time profiles as described above.
- Translational research: The ability to integrate your in vitro, small animal, and large animal compound data into a single platform facilitates translational research to better inform program advancement decisions, including experimental design, analyte selection, and timing of sampling.
- In vitro extrapolation to in vivo predictions (IVIVE) of DILI hazard in multiple preclinical species and in humans
- Predicted DILI hazard can be traced back to its mechanistic source(s)
- Incorporates inter-individual physiological variability
- Incorporates PK/PD variability
- User-friendly GUI to specify in silico experiments and visualize results
- Direct access to the underlying source code
- Transparent design and parameterization
- Inclusive incorporation of the available literature for mechanistic representation of the physiology
- Regularly updated to include leading edge science
- Developed and supported by scientific and technical teams
- Consortium members guide the development of DILIsym® and may share data to support model development and improve the collective understanding of DILI
- Informs drug development decisions and risk mitigation strategies
- Potential DILI hazard posed by specific molecules
- Impact of alternate pharmacology or experimental protocols on potential DILI hazard
- Mechanisms contributing to potential DILI hazard
- Mechanistic differences in cross-species sensitivity
- Identifies non-standard mechanistically-relevant safety biomarkers of DILI hazard
- Facilitates maximal use of data by integrating nonclinical and clinical data in a single platform
- Allows SimPops™ customization to reflect clinical population characteristics
- Rapidly realizes implications of “what if” scenarios
- User-friendly for scientists
- Customizable code for modelers