Method Profile
Use prior drug and donor information to select informative sampling times in a multicompartment gut–liver chip model.
Plan measurements and estimate intestinal permeability and gut/liver clearance without confounding them with system geometry, volume loss or donor activity.
Compartmental parent/metabolite ordinary differential equations for apical, basal and hepatic media; worked example is midazolam and 1′-hydroxymidazolam.
Structural-identifiability assessment, uncertainty propagation, dynamic sensitivity analysis and simulation–estimation verification under constrained sampling.
Prior inputs
Permeability and intestinal/hepatic clearance from existing assays; device area/volume/flow, donor activity, analytical quantification limits and uncertainty.
Reported software
DAISY for identifiability; R/RxODE simulations and ODEsensitivity analysis; Phoenix with RsNLME for fitting. The published implementation includes commercial fitting software.
Distinguish same-donor priors, activity-scaled different-donor data and unknown-donor distributions. Include sampling/evaporation volume changes. Verify selected times with simulated observations, then assess the fitted model against actual data before translation.
Midazolam worked example with five target parameters and an optimized versus shortened liver-sampling schedule.
2,500 simulated profiles; 100 selected experiments for parameter recovery, with three replicate analyses and 15% added residual variability. Reference activity datasets contained 128 hepatocyte and 24 enterocyte donors, with enzyme-specific missingness.
Optimized final liver sample at 24 hours versus eight hours, with other sampling times held unchanged.
Optimized relative parameter errors were ≤32%. Metabolite hepatic-clearance relative error increased from 32% to 75% and absolute average fold error from 1.2 to 3.2 with the shortened liver schedule.
A Roche–Manchester tutorial uses mechanistic simulations to plan gut–liver chip experiments. Its midazolam example shows that sampling sufficient for parent-drug parameters can still give poor estimates of metabolite clearance, with verification performed on simulated experiments.
