sensors Article Simultaneous Voltammetric Determination of Acetaminophen, Ascorbic Acid and Uric Acid by Use of Integrated Array of Screen-Printed Electrodes and Chemometric Tools Dionisia Ortiz-Aguayo , Marta Bonet-San-Emeterio and Manel del Valle * Sensors and Biosensors Group, Department of Chemistry, Universitat Autònoma de Barcelona, 08193 Bellaterra, Spain * Correspondence: manel.delvalle@uab.es; Tel.: +34-93-581-3235  Received: 28 June 2019; Accepted: 24 July 2019; Published: 26 July 2019  Abstract: In the present work, ternary mixtures of Acetaminophen, Ascorbic acid and Uric acid were resolved using the Electronic tongue (ET) principle and Cyclic voltammetry (CV) technique. The screen-printed integrated electrode array having differentiated response for the three oxidizable compounds was formed by Graphite, Prussian blue (PB), Cobalt (II) phthalocyanine (CoPc) and Copper oxide (II) (CuO) ink-modified carbon electrodes. A set of samples, ranging from 0 to 500 µmol·L−1 , was prepared, using a tilted (33 ) factorial design in order to build the quantitative response model. Subsequently, the model performance was evaluated with an external subset of samples defined randomly along the experimental domain. Partial Least Squares Regression (PLS) was employed to construct the quantitative model. Finally, the model successfully predicted the concentration of the three compounds with a normalized root mean square error (NRMSE) of 1.00 and 0.99 for the training and test subsets, respectively, and R2 ≥ 0.762 for the obtained vs. expected comparison graphs. In this way, a screen-printed integrated electrode platform can be successfully used for voltammetric ET applications. Keywords: electronic tongue; modifiers; acetaminophen; ascorbic acid; uric acid; partial least squares regression 1. Introduction Acetaminophen, Ascorbic acid and Uric acid (Figure 1) play an important role in humans’ life. Acetaminophen (N-acetyl-p-aminophenol or paracetamol (PA)) is an antipyretic and analgesic drug commonly used against arthritis, headache, muscle aches, menstrual cramps and fevers [1]. A high amount of PA can cause the accumulation of toxic metabolites, leading to severe and sometimes fatal hepatotoxicity and nephrotoxicity [2]. Ascorbic acid (AA) is a vitamin commonly present in many biological systems and in multivitamin formulations. It is widely employed to provide an adequate dietary intake and as an antioxidant [3]. Its excessive dose may cause headache, trouble sleeping, gastrointestinal discomfort and flushing of the skin [4]. Uric acid (UA) is the primary product of purine metabolism [5]. Continuous monitoring of UA in the body fluid is essential since its abnormal concentration levels lead to several diseases, such as hyperuricemia and gout [6]. Other diseases, such as leukemia and pneumonia are also associated with enhanced urate levels. Sensors 2019, 19, 3286; doi:10.3390/s19153286 www.mdpi.com/journal/sensors Sensors 2019, 19, 3286 2 of 13 Sensors 2019, 19, x FOR PEER REVIEW 2 of 13 1. Chemical Figure 1. Chemicalstructure structureofofthree three studied studied compounds compounds in this in this report report (Acetaminophen, (Acetaminophen, Ascorbic Ascorbic acid acid Uric and and acid). Uric acid). Several analytical methods for individual or simultaneous determination determination of of PA, AA and UA have been reported in the literature such as spectrofluorometry [7,8], spectrophotometry spectrophotometry [9,10], chromatography chromatography[11,12], [11,12],and and capillary capillary zonezoneelectrophoresis electrophoresis [13,14]. The problem [13,14]. The problemis that theseis that methods these can be expensive methods can be and need complex expensive and need procedures. complexFor these reasons, procedures. For the development these reasons, the of rapid, developmentcheap and of effective determination procedures is needed. One proposal to rapid, cheap and effective determination procedures is needed. One proposal to overcome this canovercome this can be the development of be electrochemical the development sensors [15]. This kind of electrochemical of devices sensors [15].provide This kind some advantages, of devices providesuchsome as, low detection advantages, limits, such as, wide lowlinear detection response limits,range, wide good linearstability response and reproducibility. range, good stability and reproducibility. However, However,certain certain difficulties difficulties arisearise whenwhen the simultaneous the simultaneous determination of these three determination compounds of these three is attempted.is attempted. compounds The oxidation Thepeaks oxidationof PA, peaksAA of and PA, UA AA areand almost overlapping UA are on traditional almost overlapping on electrodes traditional [16], which [16], electrodes makewhich their simultaneous determination make their simultaneous highly difficult. determination highly One solution difficult. One to solve solutiontheto main solve drawback the mainisdrawback the use ofismethodsthe use of based methods on modified based onelectrodes, which havewhich modified electrodes, fascinated have many researchers fascinated due to their many researchers due simplicity, high sensitivity, to their simplicity, and low cost. high sensitivity, and low In addition, this strategy cost. In addition, this allows some improvement based on electrocatalysis, liberation strategy allows some improvement based on electrocatalysis, liberation from surface fouling andfrom surface fouling and prevention of undesirable prevention of reactions undesirable competing reactions kinetically competing with the desired kinetically withelectrode process the desired [17]. process [17]. electrode Modified electrodes [18] can be prepared by several different techniques based on adsorbing, attaching specific specific molecules molecules(e.g., (e.g.,peptides peptides[19] [19]or or complexing complexing agentsagents [20,21]) [20,21]) to the to surface the surface by self-by self-assembled assembled monolayer monolayer [22], coating [22], coating and entrapment, and entrapment, e.g., is e.g., the formis theofform of conductive conductive ink [23]. ink The[23]. last The last strategy strategy has become has become interesting interesting for electrochemists for electrochemists in recent in recent times, times, becausethis because thisdeliberate deliberate and controlled controlled modification modification of of the electrode electrode surface can produce new surfaces with interesting properties employed for new devices and and applications applications in in electrochemistry. electrochemistry. Nature has developed and optimized Nature has developed and optimized an impressive an impressive varietyvariety of sensing of systems sensingused systems for navigation, used for spatial orientation, navigation, spatial prey detection, orientation, prey object inspection, detection, objectpeer interaction, inspection, peeretc. which provide interaction, etc. which technologist provide with inspiringwith technologist ideasinspiring for new concepts ideas forfor new sensors or improvements concepts for sensors or within the field [24,25]. improvements within Illustrating the field examples in chemicalexamples [24,25]. Illustrating sensing isinthe development chemical sensing ofiselectronic noses (EN)ofand the development electronic electronic tongues noses (EN)(ET), and both sharing electronic the concept tongues (ET), of preferring both sharinga the numberconceptof sensors (a sensor of preferring a array) number with of broad sensors selectivity (a sensorpattern, array) instead with broad of a selectivity single, highly selective pattern, sensor. instead of aThe use highly single, of this number selectiveofsensor. receptorsTheinuse a combinatorial of this number way of is what permits receptors to the animal way in a combinatorial senses is to be effective what permits in todetecting the animal thousands senses toofbe different effective compounds in detecting or situations. thousands In of the field of compounds different chemical analysis, the mainIn or situations. bioinspired the field systems of chemical take after analysis,three themammal main senses: bioinspiredsmell,systems taste and takesight. afterTherefore, three mammal there have beensmell, senses: reportedtaste electronic and sight. noses (EN) [26], Therefore, eyeshave there (EE) and beentongues reported(ET) [27]. From electronic nosesthese(EN)principles, [26], eyes (EE) the EN, and formed tongues by (ET)an[27]. array of sensors From with slightly these principles, the different EN, formed response by an to generic array compounds of sensors has been with slightly used forresponse different analysis to in generic the gas phase compounds and stands out has been its usedcloseness for analysisto artificial in the gasolfaction. phase and In the caseout stands of EE, there are to its closeness also interesting artificial advances olfaction. In thereported case of EE, in the thereliterature. are also An exampleadvances interesting is the development reported in ofthe a bioinspired literature. electronic An example white cane is the for blind people development of a using whiskers bioinspired multiple electronic sensor white cane principle for blindfor short-range people navigation using whiskers and exploration multiple [28]. for short- sensor principle rangeSimilar navigationto theand ENexploration is the ET that, [28].according to IUPAC [29], is defined as is a multisensor system, which consists Similar to ofthea EN number is theofET low-selective that, according sensors and uses to IUPAC advanced [29], is defined mathematical as is a multisensor procedures system,for signal whichprocessing consists ofbased a number on pattern recognitionsensors of low-selective and/or andmultivariate uses advanced data analysis. mathematicalThis analytical procedures systemfor applied to liquid analysis signal processing based on allows the generation pattern recognition of and/or multidimensional multivariate information data analysis. in combination This analyticalwith chemometric system applied processing, to liquidwhich analysisallows extracting allows the maximum the generation chemical information of multidimensional from these information in complex data. combination with chemometric processing, which allows extracting the maximum chemical information from these complex data. In this way, biomimetic systems, in opposition of classical approaches, use the combination of low selective and/or cross-responsive sensors to obtain rich and complementary analytical Sensors 2019, 19, 3286 3 of 13 In this way, biomimetic systems, in opposition of classical approaches, use the combination of low selective and/or cross-responsive sensors to obtain rich and complementary analytical information. Next, this complex, multi-dimensional information needs to be processed with proper data treatment tools, which is accomplished with chemometrics. This coupling has been declared one of the ways of progress in developing new sensing schemes [30]. There are different data processing tools depending on the final application needed. If this is a qualitative goal, PCA is a suitable linear visualization/pattern recognition method. This tool allows the reduction of the dimensionality of a multivariate problem and facilitates the visualization of different categories of the multivariate profiles by remarking similarities and differences between sample clusters. When the purpose is quantitative, different tools are available, given the numeric information is the end result. Some of these are Principal Component Regression (PCR), which departs from a first PCA transformation to build a multivariate regression, Partial Least Squares Regression (PLS), or Artificial Neural Networks (ANNs) [31]. In the present work, an eight sensor integrated array of screen-printed electrodes has been developed in base of a multiple screen-printed carbon electrode (SPCE) platform. The voltammetric array, consisting of Graphite/SPCE-Ink, Prussian blue/SPCE-Ink, Cobalt (II) phthalocyanine/SPCE-Ink and Copper oxide (II)/SPCE-Ink was employed for the simultaneous determination of the three aforementioned compounds (PA, AA and UA) by using the Cyclic voltammetry (CV) technique. This represents an example of resolving a mixture where heavily interfering signals are generated and resolving its components is difficulted. In other words, it is shown how to detect simultaneously the different analytes in presence of their interferents, which redox signals overlap. For showing these aspects, firstly, the behavior of the sensors was evaluated separately for each compound; secondly, peak current responses showed that all sensors had differentiated response for the three oxidizable compounds of clinical interest. Finally, a response model was developed to determine mixtures of PA, AA and UA at the µmol·L−1 level. 2. Materials and Methods 2.1. Chemicals and Reagents All solutions were made up using sterilized Milli-Q water (Millipore, Billerica, MA, USA). Cobalt (II) phthalocyanine (CoPc), Copper (II) oxide (CuO) nanopowder (<50 nm), Polypyrrole doped (PP) and Palladium, powder submicron 99.9+% (Pd), which were used as modifiers, were purchased from Sigma-Aldrich (St. Louis, MO, USA). Prussian blue was from Acros Organics (Geel, Belgium). The preparation of the ink composite was done using mesitylene and polystyrene, obtained from Sigma-Aldrich (St. Louis, MO, USA). Graphite powder (particle size < 50 µm) was received from BDH (BDH Laboratory Supplies, Poole, UK). Potassium chloride was purchased from Merck (Darmstadt, Germany). Acetaminophen (PA), Ascorbic acid (AA), Uric acid (UA) and hydrogen peroxide (H2 O2 ) solution were purchased from Sigma-Aldrich (St. Louis, MO, USA). All the measurements were carried out using 50 mM phosphate buffer (PBS) solution and 0.1 M KCl solution at pH 7. 2.2. Electronic Tongue The voltammetric ET was formed by an integrated array of eight screen-printed electrodes as working electrodes (8W110 Electrodes, ceramic substrate: 50 × 27 × 1 mm. and electric contacts composed of silver) from DropSens (Oviedo, Spain). The electrochemical cell consisted on: 8 working electrode (carbon, 2.95 mm diameter), auxiliary electrode (carbon) and pseudo reference electrode (Silver) [32]. Electrochemical measurements were performed at room temperature (25 ◦ C), using a portable Multi Potentiostat/Galvanostat µStat 8000 from DropSens controlled through its Dropview Multichannel 5.5 software package. A complete Cyclic voltammogram was recorded for each sample and for each Sensors 2019, 19, 3286 4 of 13 electrode by cyclic the potential between −1.5 and +1.5 V with a step potential of 9 mV and a scan rate Sensors 2019,−1 19, x FOR PEER REVIEW 4 of 13 of 50 mV·s Sensors . x FOR PEER REVIEW 2019, 19, 4 of 13 2.3.Modification 2.3. Modification of theElectrode Electrode Surface 2.3. Modificationofofthe the ElectrodeSurface Surface Thenanomaterial The nanomaterial SPCE/modifierwas was producedin in theform form ofaaconductive conductive ink-likecomposite. composite. The nanomaterialSPCE/modifier SPCE/modifier wasproduced produced inthe the formof of a conductiveink-like ink-like composite. The The corresponding modifier, graphite and polystyrene were thoroughly mixed with mesitylene forfor 2 h2 Thecorresponding correspondingmodifier, modifier,graphite graphite and andpolystyrene polystyrene were werethoroughly thoroughly mixed mixed with with mesitylene mesitylene for 2 h (Figure (Figure 2) using 2) using a magnetic a magnetic stirrer. stirrer. After After that, that, 2 min 2 min of sonication of sonication was was performed performed in order in order to obtain to obtain a h (Figure 2) using a magnetic stirrer. After that, 2 min of sonication was performed in order to obtain a medium medium thick thick solution. solution. TheThe ink-like ink-like composite composite was dropped was dropped 5 µLthe 5 µL onto onto the surface surface of a screen- of a screen-printed a medium thick solution. The ink-like composite was dropped 5 µL onto the surface of a screen- printed carbon carbon electrodes (SPCE) and dried ◦ at 40 °C for at least 1 h in order to remove solvent.the solvent. printedelectrodes (SPCE) and carbon electrodes driedand (SPCE) at 40 C for dried at least at 40 °C for1 hatinleast order1 htoinremove order totheremove Once the the solvent. Once was sensor the sensor wasthe prepared, prepared, next stepthe is annext step is[33,34] activation an activation in order [33,34] to enhancein order sensingto performances enhance sensingof Once the sensor was prepared, the next step is an activation [33,34] in order to enhance sensing performances modified of modified ink (Figure 3 displays ink the(Figure 3 displays typical gain achieved the typical after gain achieved activation). after activation). Electrochemical activation performances of modified ink (Figure 3 displays the typical gain achieved after activation). Electrochemical activation consisted ofcycles 10 repetitive voltammetric −1 between cycles 1.5 andat 50 mV·s −1 between 1.5 and consisted of 10 repetitive Electrochemical voltammetric activation at 50 mV·s consisted of 10 repetitive voltammetric cycles at−1.5 V using 50 mV·s 10 mM1.5 −1 between H2and O2 −1.5 in V using phosphate 10 mM H 2O2 in phosphate buffer (pH 7). After activation, electrodes were rinsed with −1.5 V using buffer 10 mM(pH 7). H 2O After 2 in activation, phosphate electrodes buffer (pH 7). were After rinsed with electrodes activation, deionized were waterrinsed and dried with deionized in air. water and dried in air. deionized water and dried in air. Figure2.2.Scheme Scheme of theexperimental experimental procedureforfor thethe electrode surface modification. Firstly, an ink- Figure 2. Schemeofofthe Figure the experimental procedure procedure for electrode the surface electrode modification. surface Firstly, modification. an ink-like Firstly, an ink- like solution solution was was was prepared prepared incorporating incorporating the the the corresponding corresponding modifier. modifier. Then, Then, 5 5 µL µL was was dropped onthe the like solution prepared incorporating corresponding modifier. Then, 5 µL wasdropped droppedon on the electrodesurface electrode surface and dried a 40a ◦40 C. °C. An activation step step is done with hydrogen peroxide torecord finally electrode surfaceand dried and dried a 40 An activation °C. An activation is done step with hydrogen is done peroxide with hydrogen to finally peroxide to finally record the the voltammetric voltammetric signal. signal. signal. record the voltammetric Figure3.3.Comparison Figure Comparisonof ofthe theCobalt Cobalt(II) (II)phthalocyanine/SPCE-Ink phthalocyanine/SPCE-Inkbeforebefore(scan (scan1) 1)and andafter after(scan (scan10) 10) Figure 3. Comparison of the Cobalt (II) phthalocyanine/SPCE-Ink before (scan 1) and after (scan 10) theactivation. the activation.Electrochemical Electrochemicalactivation activationconsists consistsofof10 10repetitive repetitivevoltammetric voltammetriccycles cyclesatat5050mV·s −1−1 mV·s the activation. Electrochemical activation consists of 10 repetitive voltammetric cycles at 50 mV·s−1 between1.5 between 1.5and and−1.5 −1.5VVusing using10 10mMmMHH 2O 2O inphosphate 22in phosphatebuffer buffer(pH (pH7). 7).After Afteractivation, activation,electrodes electrodes between 1.5 and −1.5 V using 10 mM H 2O2 in phosphate buffer (pH 7). After activation, electrodes wererinsed were rinsedwith withdeionized deionizedwater waterand anddried driedininair. air. were rinsed with deionized water and dried in air. Sensors 2019, 19, x FOR PEER REVIEW 5 of 13 2.4. Characterization by Scanning Electron Microscopy Sensors 2019, 19, 3286 5 of 13 The morphological characterization of the modified screen-printed electrode was performed by Scanning Electron Microscopy (SEM). A scanning electron microscope with field emission gun 2.4. Characterization (FEG-SEM) of Zeiss,by Scanning model Electron MERLIN Microscopy SM0087 was used. The morphological characterization of the modified screen-printed electrode was performed 2.5. Sample Preparation by Scanning Electron Microscopy (SEM). A scanning electron microscope with field emission gun (FEG-SEM) Accordingof Zeiss, to themodel MERLIN European SM0087 was[35] Pharmacopoeia used. the size of the data set needed for building the calibration is dependent on interfering properties and the number of analytes that needs to be 2.5. Sample Preparation handled in the model. In the majority of the cases, the size of the learning data set for calibration needsAccording to be largetowhen the interfering the European variations [35] Pharmacopoeia are acquired the size ofrandomly. the data set However, needed when the major for building the interferences can be controlled they can be varied according to a statistical experimental calibration is dependent on interfering properties and the number of analytes that needs to be handled design. in theInmodel. this case, In the the second majority option of thewas accomplished cases, using the size of the a tilteddata learning factorial experimental set for designto[36] calibration needs be 3large 3 (27 samples) for the train subset. This tilted model consisted of a factorial design with a 45° rotation when the interfering variations are acquired randomly. However, when the major interferences in caneach axis. Withthey be controlled this approach can be variedit is according possible totoavoid the repetition a statistical experimentalof numeric values. Meanwhile, design. the validation of the In this case, the second constructed option model was done with was accomplished an external using test set (12 a tilted factorial samples), these experimental designwere [36] 3 distributed randomly within the experimental domain (0 to 500 µmol·L ◦ −1) for each compound (Figure 3 (27 samples) for the train subset. This tilted model consisted of a factorial design with a 45 rotation 4). Samples in each were this axis. With prepared approachin buffer solutionto it is possible (50 mM the avoid phosphate repetition buffer solutionvalues. of numeric at pH 7Meanwhile, containing 0.1 theM KCl). Fresh validation stock of the solutions model constructed of pharmaceutical was done withcompounds an external weretestprepared the same these set (12 samples), day ofwere the measurements, in order to avoid/reduce the day-to-day variability. −1 distributed randomly within the experimental domain (0 to 500 µmol·L ) for each compound (Figure 4). Figure Figure 4. Representation Representation of of the the tilted tilted factorial factorial experimental experimental design design employed employed (3 (333)) where where itit can can be be observed how the test samples (in white) are distributed covering all the space of the training observed how the test samples (in white) are distributed covering all the space of the training samplessamples (in black). Samples 2.6. Data were prepared in buffer solution (50 mM phosphate buffer solution at pH 7 containing Processing 0.1 M KCl). Fresh stock solutions of pharmaceutical compounds were prepared the same day of the The statistical treatment and data analysis were performed using routines developed by the measurements, in order to avoid/reduce the day-to-day variability. authors using MATLAB R2017a (MathWorks, Natick, MA, USA); in particular, the functionalities “plsregress” from the Statistics and Machine Learning Toolbox, was the one employed for the 2.6. Data Processing response model; the web page Clustvis [37] was the one used for PCA calculation; Sigmaplot (Systat The statistical treatment and data analysis were performed using routines developed by the Software Inc., San Jose, CA, USA) was used to graphically represent and analyze the results. authors using MATLAB R2017a (MathWorks, Natick, MA, USA); in particular, the functionalities “plsregress” from the Statistics and Machine Learning Toolbox, was the one employed for the response model; the web page Clustvis [37] was the one used for PCA calculation; Sigmaplot (Systat Software Inc., San Jose, CA, USA) was used to graphically represent and analyze the results. Sensors 2019, 19, 3286 6 of 13 Sensors 2019, 19, x FOR PEER REVIEW 6 of 13 3. Results and Discussion 3. Results and Discussion 3.1. Characterization of the Surface 3.1. Characterization of the Surface A SEM characterization was performed in order to investigate the spatial distribution of the A SEM characterization was performed in order to investigate the spatial distribution of the ink- ink-nanoparticles and to verify if the particles were all on the external surface or in the inner layers. nanoparticles and to verify if the particles were all on the external surface or in the inner layers. As As can be observed in Figure 5, the different modifiers are distributed quasi-homogeneous between the can be observed in Figure 5, the different modifiers are distributed quasi-homogeneous between the graphite layers. The size of the nanoparticles is below 1 µm. graphite layers. The size of the nanoparticles is below 1 µm. Figure5. Figure 5. Scanning Scanning Electron Electron Microscopy Microscopy (SEM) (SEM) characterization characterization of of (A) (A) Graphite/SPCE-Ink, Graphite/SPCE-Ink,(B) (B) Cobalt Cobalt (II) phthalocyanine/SPCE-Ink, (C) Copper oxide (II)/SPCE-Ink and (D) Prussian blue/SPCE-Ink. (II) phthalocyanine/SPCE-Ink, (C) Copper oxide (II)/SPCE-Ink and (D) Prussian blue/SPCE-Ink. 3.2. 3.2. Voltammetric Voltammetric Array Array Response Response The The four four aforementioned aforementioned modifications modifications used used to to perform perform this this work, work, were were selected selected among among six six modifications modifications candidates candidates (Graphite, (Graphite, Cobalt(II) Cobalt(II) phthalocyanine, phthalocyanine, Copper oxide (II), Prussian Prussian blue, blue, Polypyrrole Polypyrrole dopeddoped and and Palladium) Palladium) to to construct construct the the sensor sensor array. array. This This selection selection facilitates facilitates thethe modification modification in in form form of of an an ink. ink. In addition, these nanomaterials are the the ones ones with with extended extended experience experience in in the the laboratory laboratory (although (although usedused in in aa different different electrode electrode format, format, thethe epoxy-graphite epoxy-graphite composites composites [38]). [38]). The behaviors of these sensors for each compound were evaluated individually. Once Oncethe thevoltammograms voltammogramswere werecollected, collected,a apreliminary preliminary qualitative qualitativeanalysis was analysis wasperformed performed in order in orderto evaluate thethe to evaluate complementary complementary of of thethe sensors. sensors. The Thechemometric chemometrictool toolused usedwas wasPCA. PCA.The Theinformation information collected collectedininthis case this to perform case to performthe PCA the calculation PCA calculation was thewas unfolded data. If electrodes the unfolded are redundant data. If electrodes are they would appear redundant they would superimposed in the PCA graph, appear superimposed in thewhile PCAa different graph, whileresponse will manifest a different response in their will separation. manifest in Astheircanseparation. be seen in AsFigure can 6, beeach seensensor in Figureshowed performance 6, each sensor showed in different regions. performance in This fact different accomplishes regions. This afactrelevant role in the accomplishes execution a relevant roleof in thethe electronic executiontongue, of the justifying electronicthe good justifying tongue, contribution the of the contribution good four preparedofelectrodes in the sensor the four prepared array. In electrodes in addition, the sensorthis analysis array. made us In addition, discard this analysisfrom the made sensor arrayfrom us discard system the Palladium the sensor and Polypyrrole array system the Palladium sensors, because they and Polypyrrole were not sensors, able tothey because provide were distinction not able to between the different provide distinction substances between (Figure 6). the different According substances to these (Figure 6).criteria, Accordingthe to other thesemodifiers criteria, could be classified the other modifiersascould goodbe candidates classified because as good they provided candidates a wide because range they of variability provided a wide between range of the substances. variability between In addition, the vastIn the substances. majority addition, of the signals were far vast majority away of the from were signals zero, far meaning they away from provide useful information zero, meaning they providetousefulthe system. information to the system. Sensors 2019, 19, 3286 7 of 13 Sensors 2019, 19, x FOR PEER REVIEW 7 of 13 Sensors 2019, 19, x FOR PEER REVIEW 7 of 13 Figure6. 6. Scoreplot plot of thetwotwo components obtained obtained after after Principal PrincipalComponent ComponentAnalysis (PCA) Figure Figure 6.Score Score plotofofthe the twocomponents components obtained after Principal Component Analysis Analysis (PCA) (PCA) analysis.Five analysis. Fivereplicates replicates for for each each sensor sensor were were done done determining determining the thethree threecompounds compoundsof of interest: interest: analysis. Five replicates for each sensor were done determining the three compounds of interest: Acetaminophen, Acetaminophen, Ascorbicacid Ascorbic acid andUric Uric acid. Acetaminophen, Ascorbic acidandand Uricacid. acid. After After thisthis pre-selection pre-selection step,step, voltammograms voltammograms for of for each each of the selected the selected electrodeselectrodes towards towards individual After this pre-selection step, voltammograms for each of the selected electrodes towards individual compounds compounds were secondly evaluated. Two scans were performed choosing the second individualwere secondlywere compounds evaluated. secondlyTwo scans were evaluated. performed Two scans werechoosing the second performed onethe choosing to represent second one theone to represent voltammetric the voltammetric response. to representresponse. the voltammetric response. Therefore, following the conditions described in Section 2.1, a stock solution of 300 µmol·L−1 of −1 of −1 Therefore, Therefore, following followingthetheconditions conditions described described in inSection Section2.1, 2.1, a stock a stock solution solution of 300 of 300 µmol·L µmol·L these compounds was evaluated. As can be observed in Figure 7, slightly different signals were of these compounds compounds was was evaluated. evaluated.As Ascan canbebeobserved observedininFigure Figure7,7,slightly slightlydifferent differentsignals signals were were obtained for the different compounds of interest, a necessary requirement for the performance of the obtained obtained forfor thethe differentcompounds different compoundsofofinterest, interest, aa necessary necessary requirement requirementfor forthe theperformance performance of of thethe electronic tongue. electronic electronic tongue. tongue. Figure Figure 7. Voltammetric 7. Voltammetric response response forfor Acetaminophen Acetaminophen (PA), (PA), Ascorbic Ascorbic acidacid (AA) (AA) andand UricUric acidacid (UA)(UA) using Figure 7. Voltammetric response for Acetaminophen (PA), Ascorbic acid (AA) and Uric acid (UA) using theusing the four the four finally finally selected selected inks. inks. (A) (A) CobaltCobalt (II) (II) phthalocyanine/SPCE-Ink; phthalocyanine/SPCE-Ink; (B)(B) Prussian Prussian blue/SPCE- blue/SPCE-Ink; four finally selected inks. (A) Cobalt (II) phthalocyanine/SPCE-Ink; (B) Prussian blue/SPCE- Ink; (C) Graphite/SPCE-Ink; (D) Copper oxide (II)/SPCE-Ink. The range of potential was from −1.5 to (C)Ink; Graphite/SPCE-Ink; (D) Copper (C) Graphite/SPCE-Ink; oxide oxide (D) Copper (II)/SPCE-Ink. The range (II)/SPCE-Ink. of potential The range was was of potential −1.5−1.5 fromfrom to 1.5 to V. The scan rate was 50 mV·s−1 and step rate of 9 mV. A 300 µmol·L−1 individual solution was employed for the four modified screen-printed electrodes. Sensors 2019, 19, x FOR PEER REVIEW 8 of 13 1.5 V. The scan rate was 50 mV·s−1 and step rate of 9 mV. A 300 µmol·L−1 individual solution was Sensorsemployed 2019, 19, 3286 for the four modified screen-printed electrodes. 8 of 13 3.3. Characterization of the Modified Integrated Screen-Printed Electrodes 3.3. Characterization of the Modified Integrated Screen-Printed Electrodes 3.3.1. Calibration 3.3.1. Calibration Curves Curves Toevaluate To evaluatethe thebehavior behaviorofofeach eachsensor sensorfor for each each compound compound separately, separately,some somecalibrations calibrationscurves curves were performed using Cyclic voltammetry (CV) technique representing were performed using Cyclic voltammetry (CV) technique representing the peak height which the peak height which corresponds to the maximum of the oxidation signal. This kind of experiment corresponds to the maximum of the oxidation signal. This kind of experiment is important to is important to determinethe determine thelinear linear range range and and the maximum the maximum concentration concentration of each compound of each compound for the finalfor the final experiment experiment (electronic (electronic tongue). tongue). Ascan As canbebe observed observedin in Figure Figure 8, 8, the the studied studied ranges ranges were were linear linear for for all all the the compounds. compounds. Therefore, Therefore, for the ET development the concentration working range was from 0 to −1 for the ET development the concentration working range was from 0 to 500 µmol·L for Acetaminophen, 500 µmol·L −1 for Acetaminophen, Ascorbic acid andAscorbic Uric acid.acid Theand linearUric acid. The linear relationship relationship and the calibration and the calibration equations equations for each sensor are for each sensor are represented in Table 1. represented in Table 1. Figure8.8.Calibration Figure Calibrationcurves curvesfor forthree threereplicates replicates (n(n = 3) = 3) to to determine determine thethe concentration concentration working working range range for for three the the compounds three compounds under under study for study the fourfor the four modified modified screen-printed screen-printed electrodes. (A) electrodes. (A) Acetaminophen, Acetaminophen, (B) Ascorbic acid, (B) (C) Ascorbic Uric acid.acid, (C) Uric acid. Table1.1.Calibration Table Calibrationdata data(y(y x) x) vs.vs. forfor thethe separate determination separate of Acetaminophen, determination Ascorbic of Acetaminophen, acid and Ascorbic acid Uric and acid Uric employing the integrated acid employing sensor the integrated arrayarray sensor chosen. chosen. Compounds Graphite Cobalt Cobalt (II) (II) Phthalocyanine Copper Oxide (II) Prussian Blue Compounds Graphite Copper Oxide (II) Prussian Blue y = 0.1234x + 2.6656 Phthalocyanine y = 0.1749x − 4.0453 y = 0.1696x + 7.5897 Acetaminophen y = 0.1057x + 2.7753 R2 = 0.991 Ry2 ==0.993 0.1234x + y = 0.1057x + 2.7753 R2 = 0.997 y = 0.1749x − R2 = 0.984 y = 0.1696x + Acetaminophen y = 0.0398x + 26.5307 Ascorbic Acid 2.6656 R = 0.993 R 2 = 0.991 y = 0.0311x + 8.5678 R = 0.993 2 y = 4.0453 R 2 = 0.0372x + 3.8868 0.997 y = 7.5897 R2 = 0.984 0.0214x + 9.0457 R2 = 0.997 R2 = 0.992 R2 = 0.992 y = 0.0398x + y = 0.0311x + 8.5678 y = 0.0372x + y = 0.0214x + Ascorbic Acid y = 0.1298x + 24.6275 y = 0.1569x + 15.231 y = 0.1074x + 7.6534 Uric Acid 6.5307 R = 0.997 y = 0.0534x R2+ =9.3828 0.993R2 = 0.996 3.8868 R2 = 0.992 9.0457 R2 = 0.992 R2 = 0.999 R2 = 0.988 R2 = 0.995 y = 0.1298x + y = 0.0534x + 9.3828 y = 0.1569x + y = 0.1074x + Uric Acid 4.6275 R2 = 0.999 R2 = 0.996 15.231 R2 = 0.988 7.6534 R2 = 0.995 3.3.2. Stability and Reproducibility Studies 3.3.2. After the and Stability calibration characterization, Reproducibility Studies a stability and reproducibility study was done in order to verify that the sensors were capable of supporting the number of measurements necessary for the After the calibration characterization, a stability and reproducibility study was done in order to Electronic tongue (ET) development. The procedure used to analyze the durability/stability of the verify that the sensors were capable of supporting the number of measurements necessary for the sensors consisted on measuring a stock solution of Acetaminophen (165 µmol·L−1) 30 times. A blank, Electronic tongue (ET) development. The procedure used to analyze the durability/stability of the in PBS solution, was inserted between each measurement to evaluate if the system was presenting sensors consisted on measuring a stock solution of Acetaminophen (165 µmol·L−1 ) 30 times. A blank, fouling effect. in PBS solution, was inserted between each measurement to evaluate if the system was presenting In all cases, the four sensors showed stable responses with Relative Standard Deviation (%RSD) fouling effect. of 3.9%, 6.3%, 5.5% and 8.2% for Graphite, Copper oxide (II), Prussian blue and Cobalt (II) In all cases, the four sensors showed stable responses with Relative Standard Deviation (%RSD) of phthalocyanine electrodes, respectively. No fouling effect was either observed in this study, when 3.9%, 6.3%, 5.5% and 8.2% for Graphite, Copper oxide (II), Prussian blue and Cobalt (II) phthalocyanine examining any trend among the blanks. electrodes, respectively. No fouling effect was either observed in this study, when examining any trend among the blanks. Sensors 2019, 19, 3286 9 of 13 Sensors 2019, 19, x FOR PEER REVIEW 9 of 13 The Thenext nextstep stepwas wastotostudy studythe thereproducibility reproducibilityofofconstruction constructionofofthe theink-modified ink-modifiedSPCE SPCEsensor. sensor. The experiment was done preparing four sensors by triplicate (n = 3) and measuring The experiment was done preparing four sensors by triplicate (n = 3) and measuring consecutively consecutively with withananacetaminophen acetaminophenstock stocksolution. solution.TheTheresults resultsfor foreach eachsensor sensorpresent presentaagood goodreproducibility, reproducibility, showing showingthe thebest bestfor the for thePrussian Prussian blue bluewith a relative with standard a relative standard deviation deviation (RSD) ofof (RSD) 0.8% 0.8%(Table 2).2). (Table Table 2. 2.Reproducibility Table Reproducibilityofofconstruction constructionofofeach eachsensor sensorwith withthe theresults resultsofofthe therelative relativestandard standard deviation (RSD). deviation (RSD). Sensor Sensor RSD% RSD% Graphite/SPCE-Ink Graphite/SPCE-Ink 2.92.9 Cobalt (II) (II) Cobalt phthalocyanine/SPCE-Ink phthalocyanine/SPCE-Ink 7.57.5 Copper Copper oxide (II)/SPCE-Ink oxide (II)/SPCE-Ink 1.31.3 Prussian blue/SPCE-Ink 0.8 Prussian blue/SPCE-Ink 0.8 3.4. 3.4.Qualitative QualitativeAnalysis: Analysis:Principal PrincipalComponent ComponentAnalysis Analysis(PCA) (PCA) Once Oncethethecharacterization characterizationwaswasaccomplished, accomplished,a aPCA PCAwaswasperformed performedininorder ordertotoevaluate evaluatethethe discrimination power of the sensors (Figure 9). The information collected in this case to perform discrimination power of the sensors (Figure 9). The information collected in this case to perform the the PCA PCAcalculation calculationwas the was thesensitivity ofofthe sensitivity thecalibration curves. calibration curves.Once Onceititwas wasconfirmed confirmedthat thatthe thedifferent different electrodes presented a differentiated electrochemical behavior towards the different compounds electrodes presented a differentiated electrochemical behavior towards the different compounds under study, underallowing the detection study, allowing for the three the detection for thecompounds, the next the three compounds, stepnext wasstep the was characterization of the the characterization sensor array chosen. of the sensor array chosen. Figure Figure 9.9.Score Scoreplot plotofofthe thetwo twocomponents componentsobtained obtained after after PCA PCA analysis.The analysis. ThePCA PCAshows showsthree three clusters clusters across across the the PC1 PC1 (94.3%). (94.3%). These These clusters clusters correspond correspond to to thethe studied studied compounds compounds n =n 3= replicates. forfor 3 replicates. These These results results were were obtained obtained using using thethe sensor sensor array array chosen. chosen. 3.5. 3.5.Quantitative QuantitativeAnalysis: Analysis:Partial PartialLeast LeastSquares Squares(PLS) (PLS)Regression Regression Once Oncethe thequalitative qualitativeanalysis analysiswaswascompleted, completed,the thedata datawaswasprocessed processedininorder ordertotobuild builda amodel model able able to quantify each compound individually. As results of the high complexity of thedata, to quantify each compound individually. As results of the high complexity of the data,it itwas was mandatory mandatorytotouse usepretreatments pretreatmentsthat thatleads leadstotoless lessnoisy noisyand andmore morehomogeneous homogeneousdata datainterpretation. interpretation. Eventually, Eventually, the theStandard Standard Normal Normal Variate Variate (SNV) (SNV) [39] method [39] method was was used usedasas aapretreatment pretreatment tool tooland andthethe Partial Partial Least Squares (PLS) technique for the model building. SNV method allowed to reducethe Least Squares (PLS) technique for the model building. SNV method allowed to reduce the scatter scattereffect of the effect of measurements the measurements applying an easyan applying mathematical process; itprocess; easy mathematical consists on it the subtraction consists on the ofsubtraction the measurement mean to the measure and followed by dividing the data by of the measurement mean to the measure and followed by dividing the data by its its standard deviation, obtaining standardindeviation, this way aobtaining normalized baseway in this lineafor all the samples. normalized A PLS1 base line for allapproach wasAnext the samples. PLS1employed, approach inwas which one model with single output was built for each compound. next employed, in which one model with single output was built for each compound. The number of the latent The variables (LVs) [40] for each model were also optimized to achieve the lowest error number of the latent variables (LVs) [40] for each model were also optimized to achieve the lowest possible and to avoid the overfitting. The final PLS models were defined for eight components or latent error possible and to avoid the overfitting. The final PLS models were defined for eight components variables (LVs) for orAcetaminophen latent variables and(LVs) 7 LVs for the other two. for Acetaminophen and 7 LVs for the other two. Sensors 2019, 19, 3286 10 of 13 Sensors 2019, 19, x FOR PEER REVIEW 10 of 13 In Figure 10 it can be observed the comparison of the obtained vs. expected results for training In Figure 10 it can be observed the comparison of the obtained vs. expected results for training sample set (dark dots), for each compound. In the same graph for each compound the testing sample sample set (dark dots), for each compound. In the same graph for each compound the testing sample set (white dots) were projected, allowing the visualization of the feasibility of each model. set (white dots) were projected, allowing the visualization of the feasibility of each model. Figure10. Figure 10.Obtained Obtainedvs. vs.expected expectedconcentrations concentrationsresults resultsplots plotsfor forthe thetraining trainingset set(black (blackdots) dots)and andthe the testing set (white dots) for (A) Acetaminophen, (B) Ascorbic acid and (C) Uric acid. testing set (white dots) for (A) Acetaminophen, (B) Ascorbic acid and (C) Uric acid. Thecomparison The comparisongeneral generaltrend trendwaswas expressed expressed asas thethe linear linear regression regression of the of the comparison comparison line. line. As As it it can can be be observed observed in Table in Table 3, for 3, for allall thethe studied studied compounds, compounds, the the y-interceptand y-intercept andslope slopeofofthe thetraining training andtesting and testingtest testregressions regressionsinclude include00and and11respectively, respectively,taking takinginto intoaccount accounttheir theirconfidence confidenceinterval interval (95%). Regarding (95%). Regarding the correlation correlation coefficients coefficientsforforall allthe theregressions, regressions,they areare they close to to close 1. Therefore, this 1. Therefore, satisfactory trend confirms the potential of the proposed approach. this satisfactory trend confirms the potential of the proposed approach. To evaluate also the fitting degree of the models the NRMSE parameter, Normalized Root Mean Square TableError, was calculated 3. Results according of the fitted regressiontocurves Equation (1). for obtained vs. expected values, for the training and testing subsets of samples and the three considered compounds (intervals calculated at the 95% ∑ confidence level). NRMSE: normalized root mean square error. (1) NRMSE = Intercept Set Analyte R2 Slope −1 ) NRMSE (µmol·L where xexpected is the theoretical concentration of the sample, xpredicted is the predicted concentration, j is Acetaminophen 0.962 1.00 ± 0.09 0 ± 25 the number of analytes considered, N the number of samples, cmax is the maximum concentration0.90 and Training Set Ascorbic acid 0.955 1.00 ± 0.09 0 ± 25 cmin is the minimum concentration. All the information about the models are summarized 0.97 in Table 3. (n = 27) Uric acid 0.940 1.00 ± 0.11 0 ± 31 1.12 Table 3. Results of theAcetaminophen 1.02 ± vs. 0.915 for obtained fitted regression curves −13values, 0.22expected ± 28 for the 0.7 training and Testing Set Ascorbic acid 0.762 1.07 ± 0.42 −3 ± 54 1.41 testing subsets of samples and the three considered compounds (intervals calculated at the 95% (n = 12) Uric acid 0.829 1.04 ± 0.33 −32 ± 36 0.85 confidence level). NRMSE: normalized root mean square error. Set Analyte R2 Slope Intercept (µmol·L−1) NRMSE To evaluate also the fitting degree of the models the NRMSE parameter, Normalized Root Mean Acetaminophen 0.962 1.00 ± 0.09 0 ± 25 0.90 Square Error, was calculated according to Equation (1). Training Set Ascorbic acid 0.955 1.00 ± 0.09 0 ± 25 0.97 (n = 27) Uric acid 0.940r P1.00 ± 0.11 0 ± 31 1.12 2 Acetaminophen i (xexpected 0.915 1.02 ± 0.22 −xpredicted ) −13 ± 28 0.7 j·N−1 Testing Set Ascorbic acid NRMSE 0.762 1.07 ± 0.42 = −3 ± 54 1.41 (1) (n = 12) Uric acid 0.829 1.04 cmax − cmin ± 0.33 −32 ± 36 0.85 where xexpected is the theoretical concentration of the sample, xpredicted is the predicted concentration, j is As can be observed, in practice, the test sample subset had the lowest NRMSE for the the number of analytes considered, N the number of samples, cmax is the maximum concentration and Acetaminophen and Uric acid compound, so the predictive capabilities of the models performed in cmin is the minimum concentration. All the information about the models are summarized in Table 3. satisfactory way. However, for the Ascorbic acid compound slightly larger values were obtained. As can be observed, in practice, the test sample subset had the lowest NRMSE for the Regarding to the correlation coefficients, the better values were observed for the training sample set. Acetaminophen and Uric acid compound, so the predictive capabilities of the models performed in Comparing these results with previous work [41] done in our group employing the same satisfactory way. However, for the Ascorbic acid compound slightly larger values were obtained. modifiers but with different technology, in this case, the bulk modification of a graphite-epoxy Regarding to the correlation coefficients, the better values were observed for the training sample set. composite electrode (Table 4), we can highlight as a main conclusion that the results obtained are similar, showing a slight improvement in the present report in terms of slope and intercept of the comparison lines. Accordingly, the combination of the screen-printed integrated voltammetry array Sensors 2019, 19, 3286 11 of 13 Comparing these results with previous work [41] done in our group employing the same modifiers but with different technology, in this case, the bulk modification of a graphite-epoxy composite electrode (Table 4), we can highlight as a main conclusion that the results obtained are similar, showing a slight improvement in the present report in terms of slope and intercept of the comparison lines. Accordingly, the combination of the screen-printed integrated voltammetry array sensors and chemometric tools allows the possibility to determine and quantify simultaneously a substance in the presence of other ones with overlapping redox potential. Table 4. Results of the fitted regression curves for obtained vs. expected values, for the training and testing subsets of samples and the three considered compounds (intervals calculated at the 95% confidence level) for the previous works [41]. Intercept Set Analyte R2 Slope (µmol·L−1 ) Acetaminophen 0.968 0.942 ± 0.031 32 ± 21 Training Set Ascorbic acid 0.947 0.933 ± 0.040 36 ± 25 (n = 33) Uric acid 0.923 0.873 ± 0.046 58 ± 25 Acetaminophen 0.848 0.895 ± 0.105 82 ± 71 Testing Set Ascorbic acid 0.908 0.919 ± 0.081 65 ± 41 (n = 15) Uric acid 0.753 0.871 ± 0.138 −8 ± 86 4. Conclusions The presented work reported for a first time in our group the simultaneous voltammetric detection of Acetaminophen, Ascorbic acid and Uric acid combining a multi screen-printed electrode integrated array with advanced chemometrics. This study clearly illustrates one of the capabilities of the biomimetc systems, concretely, ET. The ET strategy allowed the possibility, first to differentiate the compounds, next, to determine and quantify simultaneously a substance in the presence of other ones with overlapping redox potentials. The samples were analyzed by combining the Cyclic voltammetry (CV) technique for extracting the fingerprint of the individual substances and mixtures, coupled with chemometrics strategies, which permitted the resolution of the overlapping signal and its identification. The use of Principal Component Analysis (PCA) as qualitative tool was useful to determine the capability of the sensors to distinguish the different compounds under study, while in a further purpose resolution and quantification of ternary mixtures was achieved employing Partial Least Squares Regression (PLS) model. Therefore, this work demonstrates the advantages of screen-printed integrated electrochemical sensors for on-field analysis results in a promising methodology that could substitute the classical time-consuming, methods. Future works will try to detect these analytes in real pharmaceutical study case. Author Contributions: All authors contributed to this work. M.B.-S.-E. performed the Partial Least Squares (PLS) calculations. M.d.V. planned, got the funding and supervised the experiments. D.O.-A. performed the experiments, prepared the graphs and wrote the first version of the manuscript. Funding: This research was funded by the Spanish Government (Ministerio de Economía y Competitividad) grant number CTQ2016-80170-P. M.B.-S.-E. was funded by Secretaria d’Universitats i Recerca del Departament d’Empreses I Coneixement de la Generalitat de Catalunya and to European Social Fund, European Union through a FI fellowship. D.O.-A. was funded by Universitat Autònoma de Barcelona through a PIF fellowship. M.d.V. was funded by program ICREA Academia. Conflicts of Interest: The authors declare no conflict of interest. Sensors 2019, 19, 3286 12 of 13 References 1. Sanghavi, B.J.; Srivastava, A.K. Simultaneous voltammetric determination of acetaminophen, aspirin and caffeine using an in situ surfactant-modified multiwalled carbon nanotube paste electrode. Electrochim. Acta 2010, 55, 8638–8648. [CrossRef] 2. Madrakian, T.; Haghshenas, E.; Afkhami, A. Simultaneous determination of tyrosine, acetaminophen and ascorbic acid using gold nanoparticles/multiwalled carbon nanotube/glassy carbon electrode by differential pulse voltammetric method. Sens. Actuators B Chem. 2014, 193, 451–460. [CrossRef] 3. Hu, G.; Ma, Y.; Guo, Y.; Shao, S. Electrocatalytic oxidation and simultaneous determination of uric acid and ascorbic acid on the gold nanopartides-modified glassy carbon electrode. Electrochim. Acta 2008, 53, 6610–6615. [CrossRef] 4. Beard, E.L. The American society of health system pharmacists. JONA’s Healthc. Law Eth. Regul. 2001, 3, 78–79. [CrossRef] 5. Kaur, H.; Halliwell, B. Action of biologically-relevant oxidizing species upon uric acid. Identification of uric acid oxidation products. Chem. Biol. Interact. 1990, 73, 235–247. [CrossRef] 6. Miland, E.; Miranda Ordieres, A.J.; Tuñón Blanco, P.; Smyth, M.R.; Fágáin, C.Ó. Poly(o-Aminophenol)-Modified Bienzyme Carbon Paste Electrode for the Detection of Uric Acid. Talanta 1996, 43, 785–796. [CrossRef] 7. Dejaegher, B.; Bloomfield, M.S.; Smeyers-Verbeke, J.; Vander Heyden, Y. Validation of a Fluorimetric assay for 4-aminophenol in paracetamol formulations. Talanta 2008, 75, 258–265. [CrossRef] 8. Llorent-Martínez, E.J.; Šatínský, D.; Solich, P.; Ortega-Barrales, P.; Molina-Díaz, A. Fluorimetric SIA optosensing in pharmaceutical analysis: Determination of paracetamol. J. Pharm. Biomed. Anal. 2007, 45, 318–321. [CrossRef] 9. Sirajuddin; Khaskheli, A.R.; Shah, A.; Bhanger, M.I.; Niaz, A.; Mahesar, S. Simpler spectrophotometric assay of paracetamol in tablets and urine samples. Spectrochim. Acta Part A Mol. Biomol. Spectrosc. 2007, 68, 747–751. [CrossRef] 10. Ortega Barrales, P.; Fernández De Córdova, M.L.; Molina Díaz, A. Indirect determination of ascorbic acid by solid-phase spectrophotometry. Anal. Chim. Acta 1998, 360, 143–152. [CrossRef] 11. Akay, C.; Gümüsel, B.; Degim, T.; Tartilmis, S.; Cevheroglu, S. Simultaneous determination of acetaminophen, acetylsalicylic acid and ascorbic acid in tablet form using HPLC. Drug Metabol. Drug Interact. 1999, 15, 197–206. [CrossRef] [PubMed] 12. Thomis, R.; Roets, E.; Hoogmartens, J. Analysis of tablets containing aspirin, acetaminophen, and ascorbic acid by high-performance liquid chromatography. J. Pharm. Sci. 1984, 73, 1830–1833. [CrossRef] [PubMed] 13. Capella-Peiró, M.E.; Bose, D.; Rubert, M.F.; Esteve-Romero, J. Optimization of a capillary zone electrophoresis method by using a central composite factorial design for the determination of codeine and paracetamol in pharmaceuticals. J. Chromatogr. B Anal. Technol. Biomed. Life Sci. 2006, 839, 95–101. [CrossRef] [PubMed] 14. Sun, X.; Niu, Y.; Bi, S.; Zhang, S. Determination of ascorbic acid in individual rat hepatocyte cells based on capillary electrophoresis with electrochemiluminescence detection. Electrophoresis 2008, 29, 2918–2924. [CrossRef] [PubMed] 15. Švorc, L. Determination of caffeine: A comprehensive review on electrochemical methods. Int. J. Electrochem. Sci. 2013, 8, 5755–5773. 16. Zhu, X.; Xu, J.; Duan, X.; Lu, L.; Zhang, K.; Yu, Y.; Xing, H.; Gao, Y.; Dong, L.; Sun, H.; et al. Controlled synthesis of partially reduced graphene oxide: Enhance electrochemical determination of isoniazid with high sensitivity and stability. J. Electroanal. Chem. 2015, 757, 183–191. [CrossRef] 17. Ren, W.; Luo, H.Q.; Li, N.B. Simultaneous voltammetric measurement of ascorbic acid, epinephrine and uric acid at a glassy carbon electrode modified with caffeic acid. Biosens. Bioelectron. 2006, 21, 1086–1092. [CrossRef] 18. Bard, A.J. Chemical modification of electrodes. J. Chem. Educ. 1983, 60, 302. [CrossRef] 19. Serrano, N.; Prieto-Simón, B.; Cetó, X.; del Valle, M. Array of peptide-modified electrodes for the simultaneous determination of Pb(II), Cd(II) and Zn(II). Talanta 2014, 125, 159–166. [CrossRef] 20. Serrano, N.; González-Calabuig, A.; del Valle, M. Crown ether-modified electrodes for the simultaneous stripping voltammetric determination of Cd(II), Pb(II) and Cu(II). Talanta 2015, 138, 130–137. [CrossRef] Sensors 2019, 19, 3286 13 of 13 21. Pérez-Ràfols, C.; Serrano, N.; Díaz-Cruz, J.M.; Ariño, C.; Esteban, M. Penicillamine-modified sensor for the voltammetric determination of Cd(II) and Pb(II) ions in natural samples. Talanta 2015, 144, 569–573. [CrossRef] [PubMed] 22. Wawrzyniak, U.E.; Ciosek, P.; Zaborowski, M.; Liu, G.; Gooding, J.J. Gly-Gly-His immobilized on monolayer modified back-side contact miniaturized sensors for complexation of copper ions. Electroanalysis 2013, 25, 1461–1471. [CrossRef] 23. Cipri, A.; del Valle, M. Pd nanoparticles/multiwalled carbon nanotubes electrode system for voltammetric sensing of tyrosine. J. Nanosci. Nanotechnol. 2014, 14, 6692–6698. [CrossRef] [PubMed] 24. Del Valle, M. Bioinspired sensor systems. Sensors 2011, 11, 10180–10186. [CrossRef] [PubMed] 25. Toko, K. Biomimetic Sensor Technology; Cambridge University Press: Cambridge, UK, 2000. 26. Röck, F.; Barsan, N.; Weimar, U. Electronic nose: Current Status and future trends. Chem. Rev. 2008, 108, 705–725. [CrossRef] [PubMed] 27. Tahara, Y.; Toko, K. Electronic tongues—A review. IEEE Sens. J. 2013, 13, 3001–3011. [CrossRef] 28. Pallejà, T.; Tresanchez, M.; Teixidó, M.; Palacin, J. Bioinspired electronic white cane implementation based on a LIDAR, a Tri-Axial accelerometer and a tactile belt. Sensors 2010, 10, 11322–11339. [CrossRef] [PubMed] 29. Vlasov, Y.; Legin, A.; Rudnitskaya, A.; di Natale, C.; D’Amico, A. Nonspecific sensor arrays (“electronic tongue”) for chemical analysis of liquids (IUPAC technical report). Pure Appl. Chem. 2005, 77, 1965–1983. [CrossRef] 30. Lavine, B.K.; Workman, J., Jr. Chemometrics. Anal. Chem. 2002, 74, 2763–2769. [CrossRef] 31. Hanrahan, G. Computational neural networks driving complex analytical problem solving. Anal. Chem. 2010, 82, 4307–4313. [CrossRef] 32. Dropsens. 8W Screen-Printed Carbon Electrode. Available online: http://www.dropsens.com/pdfs_productos/ new_brochures/8w110.pdf (accessed on 25 July 2019). 33. Daniels, J.S.; Pourmand, N. Label-free impedance biosensors: Opportunities and challenges. Electroanalysis 2007, 19, 1239–1257. [CrossRef] [PubMed] 34. González-Sánchez, M.I.; Gómez-Monedero, B.; Agrisuelas, J.; Iniesta, J.; Valero, E. Highly activated screen-printed carbon electrodes by electrochemical treatment with hydrogen peroxide. Electrochem. Commun. 2018, 91, 36–40. [CrossRef] 35. Council of Europe. European Pharmacopoeia, 8th ed.; Directorate for the Quality of Medicines & HealthCare: Strasbourg, France, 2013. 36. Cetó, X.; Céspedes, F.; Pividori, M.I.; Gutiérrez, J.M.; del Valle, M. Resolution of phenolic antioxidant mixtures employing a voltammetric bio-electronic tongue. Analyst 2012, 137, 349–356. [CrossRef] [PubMed] 37. Metsalu, T.; Vilo, J. ClustVis: A web tool for visualizing clustering of multivariate data using principal component analysis and heatmap. Nucleic Acids Res. 2015, 43, W566–W570. [CrossRef] [PubMed] 38. Alegret, S.; Alonso, J.; Bartrolí, J.; Céspedes, F.; Martínez-Fàbregas, E.; del Valle, M. Amperometric Biosensors based on bulk-modified epoxy graphite biocomposites. Sens. Mater. 1996, 8, 147–153. 39. Gemperline, P. Practical Guide to Chemometrics, 2nd ed.; CRC press: Boca Raton, FL, USA, 2006. 40. Abdi, H. Partial least squares regression and projection on latent structure regression (PLS regression). Wiley Interdiscip. Rev. Comput. Stat. 2010, 2, 97–106. [CrossRef] 41. Gutés, A.; Calvo, D.; Céspedes, F.; del Valle, M. Automatic sequential injection analysis electronic tongue with integrated reference electrode for the determination of ascorbic acid, uric acid and paracetamol. Microchim. Acta 2007, 157, 1–6. [CrossRef] © 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).