Addressing the need for a tool that helps predict the poorly understood and sometimes fatal incidences of drug-induced liver injury, the non-profit Hamner Institutes for Health Sciences is partnering with Entelos to create a computer model of liver function in virtual patients.
The partnership, for which no financial details were made public, is set to run for two years and will combine Entelos' experience developing its PhysioLab biosimulation platform with Hamner’s knowledge of liver injury and systems biology, the partners said.
The goal is to develop a dual-species PhysioLab platform for drug-induced liver injury, or DILI, to allow scientists to predict how drugs and chemical agents might damage the liver in humans and rats, Mikhail Gishizky, Entelos' chief scientific officer, told BioInform.
Entelos said in a statement that the goal of the project is to build a "dynamic, mathematical model of liver function in virtual rodents and human patients" accounting for effects of genetic variations and other variables, such as patient gender, age, behavioral traits, and environmental factors. The partners hope that the platform will help guide development of biomarkers and assays to identify patients who are at increased risk for liver injury in response to specific drugs.
Paul Watkins, chairman of the project's advisory board, said that the Food and Drug Administration intends to use the DILI PhysioLab platform to evaluate drug safety, and that it will also be freely available to researchers.
"This is a public effort, there's nothing proprietary about it," Watkins said of the first version of the software platform, which should be available in two years' time. Watkins is the director of the Center for Drug Safety Sciences, which the Hamner and the University of North Carolina at Chapel Hill jointly established last September with $10 million in funding.
"Everything that I am involved in will be in the public domain," he said, though he noted that Entelos has the option to develop a commercial version of the system. Gishizky said that the company is "still working out" the terms of access for the platform.
Entelos said that two Food and Drug Administration scientists will join the advisory board for the project, but agency officials could not be reached for further comment.
The project is an extension of a two-year collaboration between Entelos and the FDA to create a PhysioLab platform for DILI that began in 2007 [BioInform, Aug. 10, 2007]. "This [project] builds on top of that," Gishizky said. "The FDA will continue to be an active participant in this with the Hamner and ourselves."
The dual-species aspect of the platform is a "key component" so it can bridge preclinical and clinical drug development, Gishizky said. "It will allow preclinical data to be interpreted and extrapolated into human response, which can then be confirmed by clinical trial and fed back into the platform."
When the DILI project began with the FDA, "we were not envisaging developing a virtual rat component," Gishezky said. The need for a dual-species concept, which would lead to a more "powerful" platform for DILI, arose from conversations with scientific advisors, many potential end-users including researchers in academia, the FDA, and pharmaceutical companies over the last year and a half, he said.
[ pagebreak ]
In general, he said, the dual species computational modeling approach dates back to Entelos' acquisition of Iconix Biosciences and its DrugMatrix toxicogenomics database in 2007.
Two scientists at the Hamner center are involved full-time with helping to build the model, but Gishizky did not offer details on how the project will be structured organizationally within Entelos.
The company previously created a computational approach to studying metabolism in diabetes for the American Society of Diabetes. That experience and information, Gishezky said, will be leveraged in the PhysioLab DILI project to help develop mathematical models of virtual rodents and virtual human patients to help simulate drug effects and mechanisms of liver injury.
The PhysioLab DILI platform will contain a population of rats and a population of humans, Gishizky said, so that users can use it for "translation of preclinical data into clinical response."
Rodents and humans have much in common in term of liver structure, said Gishizky, but there are important differences on a molecular level as well as "quantitative differences."
The platform needs to be tuned to human and rat commonalities and differences, he said.
"We have a lot of experience in working with dual-species representations," he said, referring to other PhysioLab platforms for other diseases including asthma, metabolic, and cardiovascular disease.
Making a Map
Entelos had already begun developing the DILI PhysioLab platform in its partnership with FDA, Gishezky said, and "progress has been made in creating an architectural map of the biological components."
The first step in that partnership, Gishizky said, was to "define what questions do you want the platform to address" and start mapping various processes requiring computational representation. "It's a multi-level type of mapping," he said.
"You develop a schematic of how the various tissues are going to be impacted, liver function, then go deeper into the liver, how the different cell types are interacting and then go deeper into how the different compartments within the cells are affecting the interaction," he said.
After the schematic was complete, the literature in the public domain was mined for information. The PhysioLab platform organizes the tasks at hand and used that structure to develop prototypical software, a knowledge management tool, Gishizky said. "You can put in references and information regarding biological processes," he said.
The knowledge management tool generates a "qualitative map" of liver functionality and how it relates to DILI, he said.
In this next development phase, Gishezky said, life scientists on the firm's scientific advisory board and other outside researchers and engineers will look at the data and the biology about, for example, hepatocytes and other cells such as Kupffer cells, and their interactions. They will "critique it and make sure we have captured all the key components that are necessary for this version 1 of the platform," Gishezky said.
The next step is converting that "qualitative map" into the mathematics necessary for the quantitative model. "That's where the engineering comes in," he said. "We're approaching that point right now."
The relationships and interactions are visualized with the software, he explained, and the team works together to map and then quantitatively frame the problem in ordinary nonlinear differential equations that represent the processes in the liver.
"We're developing virtual patients and populations of patients that represent adapters, tolerators, and susceptible individuals to various drugs," Gishizky said.
Some of these patients might be diabetic patients, for example, whose livers differ from non-diabetics also because they might be taking medications. "Those certainly impact liver function also," he said. "Virtual diabetic patients will be in the DILI platform, he said.
[ pagebreak ]
The platform will allow scientists to ask questions about new drugs as they affect the simulated liver of a diabetic patient. Even "transient signals" can be detected, Gishizky said, which might not be as obvious in an actual clinical trial with a limited number of patients.
This tool may also aid in developing biomarkers to segment the patient population indicating who should not receive the drug. The tool can "help to develop more predictive preclinical assays," Gishizky said.
Watkins said that the DILI simulation project is a "central effort" in the new Hamner-UNC center, but getting additional "stakeholders" to participate in the effort beyond the FDA and Entelos will be "very important."
One potential partner is the Drug-Induced Liver Injury Network, or DILIN, funded by the National Institute of Diabetes and Digestive and Kidney Diseases that collects data on liver damage due to prescription drugs and herbal products.
"There is no reason why that network or any other wouldn't want to fully share the latest information with us because it's a public common good [platform]," said Watkins, who also chairs the DILIN steering and genetics committees.
Watkins explained that Entelos' track record shaped his decision to ally himself with the company. "The proof is in what they have been able to deliver," he said, citing the fact that the company's cardiovascular simulation platform has been purchased by the FDA BioInform, Dec. 19, 2008] and that its in silico non-obese diabetic mouse developed with the American Diabetes Association [BioInform, June 7, 2004] is "a tremendously worthwhile model."
Under the terms of the agreement with ADA, all health professional members were given web-based access to the Entelos diabetic mouse models. "We certainly see this DILI platform is in the same vein, where the intention [is] to have a public domain platform that will be developed based on information from the public domain," Gishizky said.
The partnership with Entelos is already connected to DILIN's effort in purpose, he said, as they all are trying to help predict "why a drug can be safe for almost everybody but a patient will come along with a life-threatening sudden liver injury."
Other potential project partners include pharmaceutical companies, the European Medicines Agency, and additional research institutions around the world, Gishizky said.
As the partners try to engage others, Gishizky acknowledged that "we would appreciate the resources to contribute to the development of this [platform]."
In addition to data from the Hamner-UNC center, Watkins said he plans to use data from the UK's MRC Centre for Drug Safety Science at University of Liverpool’s School of Biomedical Sciences, which was established last fall.
The partners may also work with the Shanghai Municipal Center for Disease Control and Prevention, which is collecting DNA, blood, and urine prospectively and looking for new biomarkers for drug-induced liver injury from anti-tuberculosis treatments. Data from this effort could flow into the DILI simulation project, he said. "We will review it and decide if it goes into the model or not."
As the partners build these models, there are components that can be "switched over" for other applications, Watkins said. For example, homeostasis of the metabolite glutathione is the same in all cells. "If you work it out for the liver, you could work it out for the heart," he said.
Rare, But Deadly
Watkins said while scientists have a "very good understanding" of many aspects of liver function, "we don't understand at all why can a drug be safe for 99.9 percent of patients taking it … and then that next person, that 1,000th patient, gets the exact same drug, turns yellow and dies or needs a liver transplant," he said.
This project addresses these "rare" but "catastrophic" events, simulating the different pathways drugs take and the liver's reactions to those drugs.
As an example, Watkins noted that in his consulting work for the FDA he encountered an unnamed pharmaceutical company that had treated over 4,000 people with a drug candidate for six months. It worked "great," he said, but two patients developed liver problems. FDA asked the company to explain the data, and when it couldn't draw obvious conclusions, FDA requested that 20,000 patients be treated in a clinical trial for a year — a study that could run around $250 million and take around three years to complete, Watkins said.
FDA is "very cautious" in the wake of a number of high-profile drug safety problems in recent years, Watkins said. However, he noted that this heightened concern about safety is becoming a "major bottleneck to drug development."
Drug-induced liver injury, he said "is the number one reason why drugs fail to make it in development and why they end up undergoing regulatory actions."