Developing a vision for executing scientifically useful virtual biomedical experiments

Abstract

A virtual biomedical experiment (VBE) is a simulation of a wet-lab or clinical biomedical experiment. The goal of VBEs is to provide a scientifically useful method of in silico experimentation that can challenge concrete hypotheses and provide explanatory, mechanistic insight into the referent system. We envision virtual experimentation not as a supplement to traditional wet-lab experimentation, but rather as an essential part of the scientific method itself. The goal of this work is to lay preliminary groundwork for realizing this vision, through outlining requirements and describing agentbased models demonstrative of this vision. VBEs focus on reasoning by analogy; thus, a VBE includes model components analogous to particular relevant aspects of the referent experiment-from hypothesis formation to data analysis, and key concepts in between. We explore five exemplary categories of scientifically useful VBEs: hypothesis, experiment context, living counterparts, experiment agents, and measurements. We discuss how to develop model components of each category in the context of agent-based modeling. We highlight the importance of two overarching requirements: concreteness and modularity. Finally, we demonstrate this vision by describing an in silico liver model that partially fulfills the VBE vision and requirements, including components corresponding to each of the five VBE categories.

Utilizing virtual experiments to increase understanding of discrepancies involving in vitro-to-in vivo predictions of hepatic clearance

“…Hepatocyte objects can be identical in both systems. Simulated drug properties reflect target drug properties [5][6][7][8][9]. We achieve both objectives by adjusting parameters controlling the dynamics of drug interactions with hepatocytes and with other system components during execution.…”

“…Model execution is a discrete-event Monte Carlo (MC) simulation. The virtual experiment approach involves building, experimenting on, and iteratively refining concrete model mechanisms [6,7,16], while seeking a balance between more detailed biomimicry and the increase in computation programmed into the model systems. Concrete model mechanisms differ fundamentally from the equationbased models [17].…”

Contrasting model mechanisms of alanine aminotransferase (ALT) release from damaged and necrotic hepatocytes as an example of general biomarker mechanisms

“…Nevertheless, to facilitate reproducibility and support descriptions of method extensions and explanations of results, we provide abridged descriptions of methods reused herein. We utilize the vExperiment protocol outlined by Kirschner et al [12], as recently enhanced [8,13]. Mice employ a previously validated spatiotemporal MM that explains key early features of APAP hepatotoxicity [14].…”

“…To assign a value to some feature, we must measure it, e.g., count the number of APAP objects within all vHPCs at particular locations. Those measurements are made by agents within the framework at the end of each simulation cycle [8,13,15]. A Monte-Carlo specified interconnected directed graph, which can be different (within constraints) for each vExperiment, specifies flow paths for APAP and other Compounds (see Methods).…”

“…With additional evidence-based constraints, in conjunction with new validation targets and strong Similarity Criteria, it is straightforward to use the Iterative Refinement Protocol to alter and add MM features and then select among competing MM-based theories. In so doing, we can shrink the plausible constellation of MMs to a manageable size [8,13,38].…”