The practical value of an explicit response model in quantitative uses Biacore™ biosensors

Abstract

Suppose we aim to use data obtained by studying a biomolecular interaction system with a surface plasmon resonance (SPR) biosensor in quantifying some system feature. We assume a parametric mathematical model for biosensor response due to sources of mass such as analyte-ligand complexes. Some parameters represent interaction features, such as rate constants. Whenever we attempt to estimate parameters from data, we may obtain multiple estimates, regardless of the amount and quality of data. Inconveniently, we may be unable to distinguish between alternatives. This is problematic when alternative parameter values lead to very different predictions of system behaviour for a situation where we lack data. Anticipating this issue prior to data collection allows us to redesign the combination of planned experiments and model, replacing a certain failure to achieve our study’s aim with the possibility of success. The literature on SPR biosensors (and computational biology more generally) has paid little attention to this matter. In order to remedy this, it is appropriate to begin with a consideration of the assumed models. These are rarely specified completely, causing ambiguity that impedes scrutiny of their properties and comparison with other models. We demonstrate this by reviewing some model types seen in the Biacore™ biosensor literature. We propose to eliminate model ambiguity by providing a suitable framework for specifing models for biosensor data. This framework will aid future efforts to compose models for data arising from particular interaction mechanisms in a form that is amenable to scrutiny. We expect that the issues raised here will have relevance to the modelling of data obtained from other apparatus employed in quantifying binding behaviour.