CTSI Annual Pilot Awards to Improve the Conduct of Research

An Open Proposal Opportunity

Developing and testing mechanism-based translational hypotheses

Proposal Status: 

RationaleProblem: because of multiscale complexity, conceptual, mechanism-based, in vitro-to-in vivo mapping models are hard to falsify and can be flawed in ways that may not be obvious until challenged experimentally, which can be costly.  When the results of such experiments are equivocal or not supportive, the information needed to revise mechanistic hypotheses may be lacking.  Translation of in vitro phenomena to in vivo counterparts requires a mapping model.  Currently, mechanism-agnostic correlation models are common.  When translation is based on hypotheses about underlying mechanisms, the mapping model is almost always conceptual, often described in prose, supported by diagrams and, occasionally, mathematical models.  Solution: explicitly instantiate mechanistic mapping models in silico and explore their feasibility.  We envision concrete, computational, observable-in-action, experimentally challengeable mapping models that will evolve to become executable knowledge embodiments showing what does and does not translate under specific conditions.

The concept is illustrated in a recent paper (PMID: 20406856), “Tracing multiscale mechanisms of drug disposition in normal and diseased livers.”  We used, improved, and revalidated a multiscale in silico liver (ISL).  An ISL is an example of a new class of computational models.  We posited that changes in micromechanistic details from normal to diseased ISLs may have disease-causing, hepatic counterparts.  Together, the ISL and in silico methods represent an important step toward unraveling the complex influences of disease on drug disposition.  We demonstrated translating—morphing—a validated “normal” ISL into a “diseased” ISL.  Although ISLs are abstract software constructs, that transformation stands as a concrete, mechanistic hypothesis about what does and does not translate from one to the other.  The methods, which are new and novel, are designed to be extensible to whole organisms and, eventually, patients.  Being able to transform one validated ISL into another is important: it is evidence that the approach can be used to explore and challenge ideas about translation, making translational research more concrete. 

We envision “translational models” showing how, for example, in silico micromechanistic details are morphed between analogues of in vitro rat and human hepatocyte cultures, and in time between in vitro computational analogues and human analogues.  The morphing process will show what must be added and what is lost in translation.  A long-range goal for such morphings will be to provide an easily understood, mechanistic interpretation of how cause-effect relationships resulting from an experimental intervention in a wet-lab model are believed to manifest (or not) in a human analogue.  The expectation is that those relationships will have real world counterparts. 

Plan.  An essential precondition for achieving the above vision is to have two different, biologically related models (e.g., in vitro & in vivo) that have independently achieved validation targets.  We will focus on the above-cited ISLs and improved versions of in silico hepatocyte (ISH) cultures (PMID: 21768275): we will focus on translation of phenomena measured in hepatocyte cultures to corresponding, location dependent phenomena within hepatic lobules in rat and human livers.  We have designed the models so that hepatocytes can be exchanged: ISHs that have achieved validation targets under different culture conditions (ongoing during year one) can be plugged into an ISL (after its hepatocytes are removed).  The iterative refinement needed to reestablish whole-liver validation targets (including intralobular zonation) will provide a concrete theory of mechanistic attributes gained and lost in translation. 

Criteria and metrics for success.  1) Instantiation of quasi-autonomous ISH objects and documentation (a conference paper, e.g., the Winter Simulation Conference) that they can be transferred from one context (a simulated culture) to another (a simulated lobule) while retaining all the mechanistic features validated in vitro.  2) Produce a draft manuscript that demonstrates that an ISL with these new hepatocytes can also achieve whole liver phenomena, such as zonation of enzyme induction.  Produce a draft manuscript by year’s end for submission within the following four months. 

Approximate cost and brief justification.  Achieving 1 & 2 above will require several hundred cycles of iterative refinement (1 postdoc @ $48K working full time with Dr. Hunt) of in silico components (described in PMID: 20406856).  Simulation, publication, and meeting costs bring the total to $58K. 

Collaborators. Jackie Maher, UCSF Liver Center


The utility of this project/technology for assessment of liver targets in this ISH model. Can you comment on the generalizability/transferability of the work on this technology to other programs?

Adding to the reply above: in [Drug Devel Resh 72(2): 153–161, 2011] (URL below) we describe a vision in which analogues of different experimental systems, used by different programs, can exist within the same framework so that new knowledge about one system can be used to (automatically) update other analogues that use similar components. Clearly that is a distant vision, but there is a reasonably clear path forgetting there. http://onlinelibrary.wiley.com/doi/10.1002/ddr.20412/full

The utility of this project/technology for assessment of liver targets in this ISH model. Can you comment on the generalizability/transferability of the work on this technology to other programs?

The approach-our methods and model components-are designed to be generalized, to any other wet-lab, mammalian system used in research, and even to the same patient interacting with different forms of the same treatment (a small, in-progress project with the FDA), and that is reflected in the first presentation of these ideas in 2008 (see PMCID: PMC3041517). It proved necessary and essential to first show that the ideas can work on, specific, concrete problems, thus the liver focus. In addition to liver, we have used exactly the same methods to mechanistic insight into leukocyte behaviors, epithelial cell morphogenesis, and in vitro models of cancer. To insure that the methods are generalizable and transferable, it is necessary to be able to assemble complicated mechanisms (systems) out of simpler intuitive, easily understood, validated components that can be plugged together in different ways and in different contexts; and that it is easy to alter them for use in different contexts. Given that, it becomes feasible to explore what must be added to and taken away from an in vitro epithelial cell culture model to enable it to validate against several in vivo attributes of breast duct epithelial cells. We are not there yet (thus the application), but we are striving to move our models closer and closer to that vision.

-Can you explain in more details following terms: ‘mapping models’ and ‘mechanism-agnostic correlation models’ -It was not clear to me what the exact project will include and how that will be done. I understand that ISH and ISL models are already established. Will you be changing culture conditions and observe how ISH would change? -I think that the entire vision is great, but I was lacking practical details and how the entire process flows

I apologize for using specialized terms without adding definitions: I was striving to say too much within the 1-page limit. Mapping models (MM): the simplest MMs are scaling models. A MM is used to relate phenomena observed and measured in one experimental system (say a zebra fish) to that measured in a different system (e.g., a mouse). It is rare in experimental biology to have 1:1 quantitative correspondence for different biological phenomenon. In order for phenomena (say from a “proof-of-mechanism” experiment in cell cultures) to translate to patients, there must be a mapping model, a translational explanation (of course, there is always a risk that there is no MM: the cells in vitro simply do not use the same mechanisms as hypothesized counterparts in vivo). Often there is no MM; rather, there is a hypothesis (there is a mechanistic correspondence) that is short on details. A MM would begin to add details. When I revise the proposal, I’ll either provide a definition or avoid using MM. Mechanism-agnostic correlation models: correlations that make no claim about mechanistic linkage between the correlated phenomena. I can avoid that phrase too. Re: “It was not clear to me what the exact project will include ... Will you be changing culture conditions and observe how ISH would change?” Given the 1-page limit, I avoided details. When I revise the proposal, I’ll include statements, even though brief. Yes, ISH and ISL models are already established, but they are works-in-progress. A goal will be to validate ISH using data from at least two different in vitro experiments, replace current ISL hepatocytes, and then predict whole liver phenomena for which rat data is available. If our predictions are “off,” we will alter the ISH to achieve validation. How those hepatocyte mechanisms need to be altered (within the ISL) provide a testable hypothesis for how the in vitro hepatocyte mechanisms needed to be altered in order for the results to translate to in vivo. A simpler example would be the clearance of a drug measured in hepatocyte cultures. How do we translate that measure of clearance, first to rats, and then to humans? For an already-studied drug we can discover a morphing of ISH clearance mechanisms that enables them, when placed within ISL, to predict hepatic clearance it rats.

Commenting is closed.