Multivariate data analysis of dynamic amperometric biosensor responses from binary analyte mixtures-application of sensitivity correction algorithms

In this paper, it is demonstrated that a single-receptor biosensor can be used to quantitatively determine each analyte in binary mixtures using multivariate data analysis tools based on the dynamic responses received from flow injection peaks. Mixtures with different concentrations of two phenolic compounds, catechol and 4-chlorophenol, were measured with a graphite electrode modified with tyrosinase enzyme at an applied potential of −50 mV versus Ag/AgCl. A correction algorithm based on measurements of references in-between samples was applied to compensate for biosensor ageing as well as differences caused by deviations between biosensor preparations. After correction, the relative prediction errors with partial least squares regression (PLS-R) for catechol and 4-chlorophenol were 7.4 and 5.5%, respectively, using an analysis sequence measured on one biosensor. Additional validation mixtures of the two phenols were measured with a new biosensor, prepared with the same procedure but with a different batch of tyrosinase enzyme. Using the mixture responses for the first sensor as a calibration set in PLS-R, the relative prediction errors of the validation mixtures, after applying correction procedures, were 7.0% for catechol and 16.0% for 4-chlorophenol. These preliminary results indicate that by applying correction algorithms it could be possible to use less stable biosensors in continuous on-line measurements together with multivariate data analysis without time-consuming calibration procedures.