Optimal linear data analysis for surface plasmon resonance biosensors

Surface plasmon resonance biosensors measure the thickness or molecular con centration of a biolayer by analyzing small changes in measured reflection spectra. In this paper, we describe linear spectral analysis techniques des igned to produce measurements with the maximum possible signal-to-noise rat io. We show how, under appropriate assumptions, an optimal analysis method may be derived for measuring any system parameter, and how measurements of multiple parameters may be made independent in exchange for an decrease in signal to noise ratio. Compared to two conventional data analysis technique s (quadratic fit and centroid methods) using simulated data, the linear tec hniques show a 30% increase in signal to noise ratio. In application to act ual thiol binding data, the linear method yields a signal to noise ratio 46 % greater than that of the centroid method and 65% greater than that of the quadratic fit method. This level of noise reduction was achieved by using the ability of the linear methods to reject noise caused by light source br ightness variations. (C) 1999 Elsevier Science S.A. All rights reserved.