Abstract
Comparison of experimental data with modeling predictions is essential for making quantitative measurements of species properties, such as diffusion coefficients and species concentrations using a T-sensor. To make valid comparisons between experimental data and model predictions, it is necessary to account for uncertainty in model predictions due to uncertain values of model parameters. We present an analysis of uncertainty induced in model predictions of a T-sensor based competitive diffusion immunoassay due to uncertainty in diffusion constants, binding reaction rate constants, and inlet flow speed. We use a non-intrusive stochastic uncertainty quantification method employing polynomial chaos expansions to represent the dependence of uncertain species concentrations on the uncertainty in model parameters. Our simulations show that the uncertainties in model parameters lead to significant spatially varying uncertainty in predicted concentration. In particular, the diffusivity of fluorescently labeled probe antigen dominates the overall uncertainty. The predicted uncertainty in fluorescence intensity is minimum near the centerline of T-sensor and relatively high in the regions with gradients in fluorescence intensity. We show that using centerline fluorescence intensity instead of first derivative of fluorescence intensity as the system response for measuring sample antigen concentration in T-sensor based competitive diffusion immunoassay leads to lower uncertainty and higher detection sensitivity.