Low-cost, syringe based ion-selective electrodes for the evaluation of potassium in food products and pharmaceuticals.

• Development of low-cost ion-selective electrodes made from disposable syringes. • Standard quantification of potassium is not precise for samples of high ionic strength. • Two electrodes coupled with ensemble random forest model greatly enhance precision. • Potassium was quantified in food and pharmaceutical supplements without recalibration. • Similar performance can be achieved with lower-cost ionophores using auxiliary sensors. Over the years, ion-selective electrodes (ISEs) have become routine means of analysis for industrial, environmental, and clinical applications, allowing for rapid, non-invasive, and easy disease monitoring and diagnostics. Their enormous advantages include wide dynamic range, ease of operation and low detector cost. Sets of ion-selective electrodes selective towards various ions can easily be assembled into an electronic tongue system. The additional information gained from other electrodes can help overcome the ISE's inherent limitations, such as the need for recalibration in more complex samples of variable ionic strength. In this work we developed low-cost ion-selective electrodes made from disposable syringes equipped with plasticized PVC membranes. To characterize the system we have chosen well-studied ionophore valinomycin, and the results in model solutions were comparable to standard ISE electrodes fabricated using commercial bodies. However, quantification of potassium in real samples showed, as expected, that measurement in complex, more concentrated solutions of higher ionic strength, such as beetroot soup, tomato-based food products, and dried fruits, are reproducible but subject to relative error up to 76 %. To overcome this limitation, we have constructed an electronic tongue based on the low-cost syringe electrodes. Different compositions of the array were tested with a series of multivariate algorithms. We have shown that adding a single type of electrode coupled with an ensemble random forest model allows us to quantify potassium in a wide range of food products, and pharmaceutical supplements without recalibration. Compared to a quantification of potassium using a standard calibration curve, the root mean square error of prediction was almost six times lower. We have also shown how the array can be designed to achieve comparable analytical performance using less selective ionophores, such as dibenzo-18-crown-6 instead of valinomycin. [Display omitted] [ABSTRACT FROM AUTHOR]