Due to surface rugosity, the physical and chemical microenvironment at the surface is heterogeneous. Therefore, proteins that have a single class of binding sites in solution can show a dispersion in the binding properties, once chemically crosslinked to the surface. As a tool to study this ensemble of surface sites, we previously introduced a computational approach to determine distributions of affinity and kinetic binding site parameters from experimental data. We have now extended this approach to incorporate first-order approximations of mass transport limitation. It allows us now, for the first time, to account simultaneously for the two most commonly encountered experimental problems of surface binding when using biosensors to characterize protein interactions, site heterogeneity and mass transport limitation, and thereby model experimental data to within the level of noise of data acquisition, and fully exploit the high sensitivity of surface plasmon resonance biosensors for the study of protein interactions. ? ? We have applied this technique to the detailed study of hydrodynamic effects on analyte transport near the surface. In the presence of a polymeric immobilization matrix, the data clearly show significantly smaller mass transport coefficients as compared to the theoretical predictions from a laminar flow model. The origin of the additional, rate-limiting step is unclear, but appears to be due to hydrodynamic or electrostatic interactions of the soluble analyte with immobilized polymers resulting in reduced protein diffusion, resulting in limited permeability of the polymer matrix for soluble analytes. ? ? We have also embarked on using this new tool for functionally characterizing the ensemble of surface binding sites to study the properties of different types of surfaces, strategies for surface attachment, and effect of surface density of immobilization. As expected, first results show that these factors can significantly influence the binding properties of the surface sites.? ? Recently, we have refined the computational approach by introducing Bayesian regularization that enables us to introduce different prior expectations of the shape of the distribution. This approach helps to increase the resolution of the method, and to make detailed comparison of different distributions.
