Quantifying Assay Confidence in Barrier Organ-on-Chip Systems: A Monte Carlo Assay-Readiness Framework for Detecting Modest Permeability Shifts

Abstract

Organ-on-chip (OoC) systems are increasingly used to generate human-relevant evidence in drug discovery and preclinical development, yet interpretation can be limited by the interaction of biological variability, device-to-device variability, and technical assay noise. Here, we describe the UstarFlowAI barrier-confidence framework, an in silico assay-readiness approach that quantifies whether a predefined permeability-related effect is resolvable under specified uncertainty. The proposed context of use is decision support for laminar, low-Reynolds-number airway epithelial monolayer barrier OoC assays in which device fabrication, pump performance, flow, wall shear stress, and biological fluctuation may affect permeability-related readouts. Monte Carlo simulations propagate defined manufacturing, biological, systemic, and well-level uncertainty and estimate the minimum detectable shift (MDS) under a prespecified decision rule. In the present proof-of-concept design, the reference workflow was parameterized at 9.4% technical coefficient of variation (CV), whereas the standardized workflow was parameterized at 5.0% CV, corresponding to an approximately 46.8% reduction in the modeled technical-noise component. At n = 48 replicate wells per group and a nominal total assay-noise operating point of 17% CV, a predefined 10% permeability shift exceeded the upper MDS uncertainty bound for the standardized workflow in both best- and worst-case scenarios, whereas it overlapped the 95% MDS interval for the reference workflow and therefore did not meet the prespecified PASS criterion. These findings demonstrate the behavior of the computational framework under the stated assumptions; they do not constitute experimental validation, regulatory qualification, or evidence that a 10% shift is universally meaningful across barrier OoC systems. We therefore propose a prospective wet-lab validation strategy in which model predictions are tested against airway barrier measurements generated under controlled sources of technical and biological variability. The framework is intended to complement, rather than replace, the underlying biological model by making assay uncertainty and decision thresholds explicit and auditable.