ABSTRACT:We sought a single set of mechanisms that could provide a quantitative explanation of three pairs of published time series data: perfusate concentration of digoxin and its metabolite in perfusates of isolated perfused rat livers 1) in the absence of any predose and with a predose of either 2) the uptake inhibitor rifampicin or 3) the efflux inhibitor quinidine. We used the synthetic modeling and simulation method because it provides a means of developing a scientific, experimental approach to unraveling and understanding some of the complexities of drug-drug interactions. We plugged together validated, quasi-autonomous software components to form abstract but mechanistically realistic analogs of livers undergoing perfusion [recirculating in silico livers (RISLs)], into which we could add objects representing each of the above three drugs, alone or in combination. Each RISL was a hypothesis about plausible mechanisms responsible for the referent time series data. Simulations tested each hypothesis. We used similarity measures (SMs) to compare results to the six sets of referent data. From many candidates, we identified an RISL having time-invariant mechanisms that achieved a weak SM (SM-1) but failed to achieve a stronger SM. Replacing four time-invariant with time-variant mechanisms along with addition of new enzyme and transporter components achieved the most stringent SM: simulated digoxin and metabolite perfusate levels were experimentally indistinguishable from the referent data for all three treatments. The mechanisms simulated unanticipated loss of hepatic viability during the original wet-lab experiments: erosion of hepatic accessibility and of enzyme and transporter activities.Given the variety of transporters and enzymes that can be involved, how can we enhance substantially our ability to confidently anticipate the consequences of hepatic drug-drug interactions in advance of costly wet-lab experiments? To do so, we need improved knowledge of multilevel spatiotemporal mechanistic details. We need new classes of models with capabilities beyond those of the current pharmacokinetic (PK) and pharmacodynamic variety, plus methods to more realistically represent critical spatiotemporal details, all within singlemodel systems. Finally, we must be able to explore realistic, concurrent, interaction consequences of two or more drugs when given key physicochemical properties, which is infeasible using current PK models. An essential first step in realizing those needs is to show a fine-grained, computational analog in which measures of simulated interactions of two drugs are experimentally indistinguishable from those of referent wet-lab experiments. When the analog's assembled components map realistically to hepatic components and the measured consequences of in silico and wet-lab mechanisms are indistinguishable, we can posit that the responsible mechanisms of both systems are similar, as diagrammed in Fig. 1. Achieving these important goals has been an objective of this project.The referent wet-lab exp...