Design of Potentiometric Sensor Arrays Using Fisher Information and Genetic Algorithm

Abstract

Potentiometric sensor arrays, or electronic tongues, are based on combining cross-sensitive electrodes with multivariate chemometric methods for the simultaneous quantitative determination of analytes in complex liquid media. While cross-sensitivity is recognized as a key feature of electronic tongues, there are currently no a priori theoretical approaches to evaluate which combination of cross-sensitive potentiometric sensors can form an effective array for quantitative multi-ion analysis prior to experimental trial-and-error. In this work, we report the derivation of a Fisher Information-based objective function and its implementation with genetic algorithm for a priori sensor selection in potentiometric sensor arrays. As an illustration of the utility of our method, we demonstrate the design of a potentiometric sensor array for the quantitative determination of Na+, K+, Mg2+, and Ca2+ in blood serum through the screening of a library of more than 300 ion-selective electrode membranes. The results of our analysis suggest that array configurations which are predicted to minimize error can have complex patterns of analyte cross-sensitivities. These alternative array configurations can be difficult to deduce intuitively or to discover by experimental trial-and-error. Simulated sensor array responses modeled by artificial neural networks demonstrate the utility of our our method to rank the performances of sensor array configurations.