Session 1: Refining Concrete Cascade Networks

Refining Concrete Cascade Networks to Explain Acetaminophen Hepatotoxicity Phenomena Within and Across Multiple Scales
OpenTox Euro 2014 speaker: C. Anthony Hunt

C. Anthony Hunt


Bioengineering and Therapeutic Sciences, University of California, San Francisco


C. Anthony Hunt and Andrew K. Smith, Bioengineering and Therapeutic Sciences, University of California, San Francisco, Glen E.P. Ropella, Tempus Dictum, Inc., USA

Mon, 22. Sep. 2014


I am Professor of Bioengineering and Therapeutic Sciences, Schools of Medicine and Pharmacy, The University of California, San Francisco. I direct the BioSystems Group. A primary focus is quantitative and systems pharmacology modeling and simulation. We research, develop, and use advanced (from the perspective of biomedical research) modeling and simulation methods to achieve deeper, actionable
insight into the networked, layered, macro- and micromechanisms that link molecular level events with higher level phenomena and operating principles across scales at cell, tissue, organ, organism, and population levels in the presence and absence of interventions. Current research includes
1) the coupled influences of inter-, intracellular, and tissue heterogeneity on xenobiotic (drug) transport, metabolism, and response
in normal and diseased livers,
2) epithelial cell morphogenesis and tissue repair in vitro,
3) using simulation to individualize treatment interventions, and
4) using simulation to concretize plausible mechanisms responsible for subject-by-formulation interactions that, in turn, cause large intra- and interindividual bioavailability variability.

I am a member of the Editorial Boards of Computers in Biology and Medicine, Simulation, the International Journal of Knowledge Discovery in Bioinformatics, and the Journal of Computational Biology and Bioinformatics Research.

I am an AAAS and AAPS Fellow, and a Director of The McLeod Modeling and Simulation Network. I have consulted for various companies.

The challenge for this work has been discovering and iteratively improving minimal in silico component interactions that stand as challengeable mechanistic hypotheses capable of explaining characteristic features of APAP hepatotoxicity. To insure broader usefulness within the larger pharm/tox context, we adhere to delineated near and long-term requirements along with best software engineering, simulation, and scientific practices.

METHODS: We use discrete event, agent-based analogs comprised of nested modular spaces and components assembled into biomimetic mechanisms. Our validation targets are drawn from a diverse set of phenomena that we wish to explain eventually. Each phenomenon is a targeted attribute (TA). We cycle through a five-stage Iterative (mechanism and component) Refinement Protocol.

1) Specify the subset of TAs to be validation targets during the current work cycle along and quantitative similarity criteria (SC). A validation target is achieved when an analog phenomenon attains the SC prespecified for a TA.
2) Formulate minimal computational mechanisms intended to mimic essential features of referent mechanisms. Phenomena generated during execution are products of component interactions.
3) Instantiate those mechanisms by refactoring and reparameterizing components and mechanisms from already validated analogs. Design and execute experiments.
4) Use SC to compare corresponding simulated and wet-lab measurements. When SC for several TAs are achieved, the analog has attained a degree of validation: it stands as a plausible, concrete, explanation of the targeted phenomena.
5) Challenge/falsify Stage 4 mechanisms in the next work cycle by including new TAs at Stage 1.

RESULTS: I will describe achieving multiple validation targets, including dose dependent hepatocyte necrosis occurring first adjacent to lobule central veins. At stage 5 in the most recent cycle, we added TAs characterizing SP600125 inhibition of hepatocyte necrosis following a single dose given two hours after a toxic APAP dose. Simulating inhibition required:

1) separate SP600125 objects that, simultaneous with APAP objects, percolate through and interact with components;
2) giving analog components the ability to distinguish between APAP and SP600125 and interact appropriately; and
3) achieving SP600125 absorption, distribution, and clearance validation targets. I will describe a coarse grain interaction mechanism that enabled achieving time-dependent validation targets for inhibition of necrosis.

CONCLUSION: The evidence presented strengthens the case that biomimetic synthetic analogs using concrete parsimonious mechanistic hypotheses provide an explanation for how APAP toxicity emerges within and across biological scales, and that enhances prediction credibility. The methods facilitate in silico experimentation to falsify (or not) previously validated mechanisms.

Analogs and mechanisms are intended for reuse in studying any number of xenobiotics, alone and in combination. They make it straightforward to refine components iteratively to establish new, more explanatory mechanisms.