Discrimination of three bacteria species using a potentiometric electronic tongue
Conference Paper
Overview
Research
View All
Overview
abstract
The detection, monitoring and/or prevention of microorganism growing is of utmost relevance in several research fields, from food to environmental areas, being an important topic either from an academic or an industrial point of view. Conventional methods like plating techniques are the most widely used, being needed novel and faster screening methodologies like electronic noses, electronic tongue (E-tongue) and impedance based methods. In the present work, a potentiometric E-tongue (Fig.1), comprising 40 lipid polymeric sensor membranes with cross-sensitivity, was used to identify and discriminate three bacteria (Enterococcus faecalis ATCC29212; Staphylococcus aureus ATCC653 and Escherichia coli ATCC29998) at two concentrations levels (low and high). Brain Heart Infusion Broth medium was used for cultivating each of the three microorganisms, which were then individually inoculated into 1L Erlenmeyer flask (working volume of 300 mL) and incubated overnight (batch mode) at 35ºC, on a rotary shaker (90 rpm). After incubation, the biomass was spectrophotometrically determined, being measured the optical density at 550 nm. The cultures were split in volumes of 50 mL. The cells of each sample were harvested (centrifugation at 9000 rpm for 10 min) and stored at -20ºC. The obtained biomass was dried at 30ºC, and stored at -20ºC. Before E-tongue analysis, the cells were rehydrated with 20 mL of deionized water for 30 minutes at room temperature and aqueous sample solutions with different cell concentrations were obtained. Each E-tongue analysis took five minutes, enabling establishing a pseudo-equilibrium between the samples and the sensors´membranes, being the signals potentiometric profiles recorded. The classification performance of E-tongue was assessed by applying a linear discriminant analysis (LDA) coupled with the meta-heuristic simulated annealing (SA) variable selection algorithm. The preliminary results showed that E-tongue-LDA-SA predicting model could be established, based on the information gathered by a sub-set of 15 sensors, allowing to correctly classify (100%) (Fig.2) and 85% (original and leave-one-out cross-validation procedure, respectively) of the samples according to the microorganism and respectively concentration level.
The microorganisms were inoculated into the broth medium and grown 90 rpm and 37º overnight in a rotatory incubator
