Sugars’ Quantifications Using a Potentiometric Electronic Tongue with Cross-Selective Sensors: Influence of an Ionic Background
, Adélio A. S. C. Machado 4, Evandro Bona 5 and Luís G. Dias 1,*,†Round 1
Reviewer 1 Report
The paper by V. da Costa Arca et al on sugars quantification from electronic tongues is well written and the conclusion are clearly supported by the results. The experiments have been performed rigorously with strong statistical analysis.
I believe that this paper deserve publication in Chemosensors but I would strongly suggest the authors to improve the introduction especially by enhancing the references on cross-reactive organic/polymeric receptors. I would suggest to add the following references:
Riul et al. Biosensors Bioelectron 2003, 18, 1365.
Miranda et al. J. Am. Chem. Soc. 2007, 129, 9856.
Hou et al. Angewandte Chemie 2012, 51, 10394.
Genua et al. Talanta 2014, 130, 49.
Huynh et al. Biosensors Bioelectron 2015, 74, 856.
Garçon et al. Sensors 2017, 17, 1046.
Wang et al. Chem. Eur.J. 2017, 23,12471.
Author Response
Reviewer 1
Changes were carried out using the color yellow
The paper by V. da Costa Arca et al on sugars quantification from electronic tongues is well written and the conclusion are clearly supported by the results. The experiments have been performed rigorously with strong statistical analysis.
I believe that this paper deserve publication in Chemosensors but I would strongly suggest the authors to improve the introduction especially by enhancing the references on cross-reactive organic/polymeric receptors.
We kindly acknowledge the Reviewer’s comment and suggestions to improve the article.
New corrections were made in the revised Manuscript and we expect that they have been according to the Reviewer’s expectations.
Page 2, lines 77-78 – old manuscript
“In this context, the performance of a potentiometric sensor array with cross-sensitivity sensors for the simultaneous analysis of three sugars (glucose, fructose and sucrose) was studied in the present work, as a fast, cost-effective and green alternative analytical tool.”
In the new new manuscript, lines 75-93, page 2, it was introduce the following text:
“In this context, electronic tongues are alternative analytical tools, consisting of sensor arrays with low-selective and nonspecific sensors coupled with chemometric techniques for signal processing. The cross-sensitivity is a typical ability of a low-selective sensor to respond reproducibly to a number of different analytes in solution [11]. The most common electronic tongue design is a sensor array composed with selective lipid/polymer membrane, chalcogenide glass, porphyrin derivatives, or conducting polymer using different signal transduction (electrochemical, spectroscopic, or gravimetric) that allows the discrimination of taste standard substances [12]. New architectures of electronic tongues using cross-reactive receptors can be found, for instance: an artificial taste sensor based on conducting polymers using a combination of both supra-molecular thin films of conducting polymers with a lipid-like material and impedance spectroscopy [13]; an sensor array based on fluorescence response sensing of proteins by using different functionalized poly(p-phenyleneethynylene)s, highly fluorescent polymers, which possess various charge characteristics and molecular scales [14]; an electronic tongue based on molecularly imprinted polymers (highly crosslinked polymer matrices formed with a cavity with shape and functional group complementary to the target analyte molecules) giving a unique pattern using several transduction platforms, from spectroscopic to electrochemical [15]; an electronic tongue by coupling surface plasmon resonance imaging with an array of non-specific and cross-reactive receptors, by mixing in different proportions two hydrophilic compounds, disulfide with lactose (neutral) and sulfated lactose (highly negatively charged), for food analysis [16].”
Due to the change of text, 7 new articles were also inserted in the new manuscript:
11 Y. Vlasov, A. Legin, A. Rudnitskaya, C. Di Natale, A. D’Amico, Nonspecific sensor arrays (“electronic tongue”) for chemical analysis of liquids (IUPAC Technical Report), Pure Appl. Chem77(2005) 1965–1983.
12 K. Toko, A taste sensor, Meas. Sci. Technol.9 (1998) 1919–1936.
13 Jr .A. Riul, R.R. Malmegrim, F.J. Fonseca, L.H.C. Mattoso, An artificial taste sensor based on conducting polymers, Biosens. Bioelectron. 18 (2003) 1365/1369.
14 O.R. Miranda, C.-C. You, R. Phillips, I.-B. Kim, P.S. Ghosh, U.H.F. Bunz, V.M. Rotello, Array-Based Sensing of Proteins Using Conjugated Polymers, J. Am. Chem. Soc. 129 (2007) 9856-9857.
15 T.-P. Huynh, W. Kutner, Molecularly imprinted polymers as recognition materials for electronic tongues, Biosens. Bioelectron. 74 (2015) 856–864.
16 M. Genua, L.-A. Garçon, V. Mounier, H. Wehry, A. Buhot, M. Billon, R. Calemczuk, D. Bonnaffé, Y. Hou, T. Livache, SPR imaging based electronic tongue via landscape images for complex mixture analysis, Talanta 130 (2014) 49–54.
Author Response File: 
Author Response.docx
Reviewer 2 Report
The submitted manuscript is very interesting and describes the potential of using an electronic tongue to quantify different sugars in solution. However, prior to considering its publication in Sensors, several issues must be addressed apart some mistyping errors:
1) Were samples with one sugar also prepared?
2) How many measurements were performed for each solution (from Table 1)?
3) What was the measuring time (per sample)? How repeatable was sensor response?
4) Authors are encouraged to provide sensor schematics in Section 2.3 (despite stating it is shown in their other publication).
5) Can the authors provide any information regarding sensor reset? Authors mention only the cleaning procedure between sample analysis (lines 182-184)
6) Figure 2 shows sensors response for all 25 samples described at Table 1 or it is from a specific sample? If so, which one?
7) Have the authors tried to build more ET systems to better understand the low repeatability of sensors’ response (Figure 2)?
8) What was the drop volume (sensor assembly) – line 168?
9) Authors refer to table 2 in line 176, right?
10) Table 4 should be in vertical alignment for better visualization. Its current form is not acceptable.
11) It is not clear to me how the authors could determine which sensor (array) were better for each sugar (results described at Table 3). If I understood correctly, all samples analyzed contained the three sugars at different ratios and the ET used is composed by non-specific sensors. So how could the authors discriminate the sugar on those solutions.
12) The same question occurs for the results presented in Figure 4.
13) Even though some differences are expected from the duplicated systems, I was surprised with the results presented at Table 3. On average, 16 sensors were the best fit for each sugar, however, only 4 to 5 sensors from ET1 were also selected from ET2. Example of selected sensors for fructose (with KCl): ET1 – 4, 8, 12, 14, 15 and 16; ET2 – 2, 3, 8, 14, 15, 16, 17, 18, 19 and 20. Common (duplicated) sensors: 8, 14, 15 and 16.
A higher number of common sensors were expected when a duplicated system is used (higher repeatability on sensors response / assembly). Could authors comment on that?
14) Based on the results described at Table 3, which sensor array should be used for sugar quantification?
15) Have the authors evaluated system performance (sugar quantification) using just the data collected with ET1 (or ET2) or both systems are needed?
Author Response
Reviewer 2
Changes were carried out using the color green.
Comments and Suggestions for Authors.
The submitted manuscript is very interesting and describes the potential of using an electronic tongue to quantify different sugars in solution. However, prior to considering its publication in Sensors, several issues must be addressed apart some mistyping errors:
We acknowledge the positive comments of the Reviewer and we would like to thank all the efforts made through the correction process, since its suggestions were a major contribution to improve the present manuscript.
1) Were samples with one sugar also prepared?
No. The sugar’s concentration range was defined considering previous studies were the electronic tongue was applied, as in honey and juice beverages analysis. The purpose of this work was to verify the possibility of the multi-sensor system to quantify each sugar in mixed solutions. Considering zero levels of concentration of a sugar would involve further assays and other multivariate modulation techniques in order to establish limits of detection and quantification.
This issue can be addressed in a future work, considering that it can give a research study with interest and novelty.
In line 130, page 3 (old manuscript), the phrase:
“The individual sugar concentrations were within the dynamic range of concentration of a HPLC method validated in a previous work [14].”
was changed to (line 143, page 4):
“The individual sugar concentrations were within the dynamic range of concentration of a HPLC method validated in a previous work for soft drinks analysis [14].”
2) How many measurements were performed for each solution (from Table 1)?
Each solution was analyzed twice in order to verify its repeatability and confirm that the signals profile was correct. This information was included in the Revised Manuscript.
In line 200, page 6, a new phrase was included:
“All solutions were measured twice to ensure that the signal profile obtained was correct.”
3) What was the measuring time (per sample)? How repeatable was sensor
response?
In the sub-chapter 2.5. Potentiometric multi-sensor device is referred that analysis was carried out for 5 min at room temperature, not considering the initial 1-2 minutes between analysis where the e-tongue was washed with deionized water, dried with absorbent paper and inserted into a new solution for analysis. Only after these steps was the data from the previous analysis saved into an excel file and then the 5 minutes period was started for analysis of the new solution.
In line 183, page 5 (old manuscript), the phrase:
“Between determinations the analytical system was flushed thoroughly with deionized water and lightly dried with absorbent paper.”
was changed to (line197, page 5):
“Between determinations (within a period of 1-2 minutes) the analytical system was flushed thoroughly with deionized water, lightly dried with absorbent paper, inserted into a new solutionfor analysis and, after these procedures, the data from the previous analysis was saved.”
The repeatability of sensors (data not included in the article) in the electronic tongue is always verified before starting a research work. To clarify this point, a new sentence has been added to page 8, line 288:
“Overall, the lipid polymeric membranes on the sensor arrays had satisfactory signal stability over time analysis (during 5 min of signal record), with relative standard deviations (%RSD) lower than 3.5%, and repeatability, with %RSD values between 0.5 and 12%.“
4) Authors are encouraged to provide sensor schematics in Section 2.3
(despite stating it is shown in their other publication).
A new figure (Figure 1) was introduced in the Revised Manuscript showing the sensor schematics of the electronic tongue. The figure was inserted in the end of the revised manuscript.
Figure 1 – Sensor schematics of the electronic tongue.
5) Can the authors provide any information regarding sensor reset? Authors
mention only the cleaning procedure between sample analysis (lines 182-184)
Regarding sensor reset, between inter-day analysis the multi-sensor system was inserted in a HCl solution (0,1 mol/L). In line 201, page 6a new phrase was included: “Between inter-day analysis the multi-sensor system was inserted in a HCl solution (0,1 mol/L).”
6) Figure 2 shows sensors response for all 25 samples described at Table 1 or
it is from a specific sample? If so, which one?
The old Figure 2, renamed as Figure 3 in the Revised Manuscript, shows the response for all solutions. In the legend of this figure was included the word “all” in order to clarify this issue, as well in the phrase on page 8 and line 294.
7) Have the authors tried to build more ET systems to better understand the
low repeatability of sensors’ response (Figure 2)?
We already prepared numerous electronic tongues with several architectures and materials. The polymeric membranes used were always the same and their performance was always verified using standard taste solutions (data not shown) that allowed us to verify their good repeatability. Our multi-sensor system is considered an electronic tongue since it allows to discriminate the five basic taste standards showing their low selectivity and cross-sensitivity (respond reproducibly to a number of different analytes) but the sample signal profile repeatability was always acceptable.
In the previous point, we already mentioned that the figure’s legend has been changed to prevent a new reader from having the same misinterpretation.
8) What was the drop volume (sensor assembly) – line 168?
The drops volume used to apply the polymeric membranes on the wells of the multi-sensor deveice was of 18 mL. This information was included in the Revised Manuscript in the line 183, page 5, by including the text “(volumes of 18 mL)”.
9) Authors refer to table 2 in line 176, right?
Correct. The text was changed to Table 2 (line 191, page 5)
10) Table 4 should be in vertical alignment for better visualization. Its current
form is not acceptable.
For better visualization, the confidence interval values were removed from Table 4, considering that these values can be calculated.
11) It is not clear to me how the authors could determine which sensor (array)
were better for each sugar (results described at Table 3). If I understood
correctly, all samples analyzed contained the three sugars at different ratios
and the ET used is composed by non-specific sensors. So how could the
authors discriminate the sugar on those solutions.
The sensors used in the electronic tongue have low selectivity and cross-sensitivity, which means that each sensor respond reproducibly to a number of different analytes. However, for the multivariate calibration model is needed to select the best sensors that contribute for each sugar quantification and remove the sensors that contribute with noise. So, the MLR model is a linear combination of the sensors’ signals selected by the simulated annealing algoritm, which implies that these sensors have information about the compound that the MLR model pretends to quantify. It is necessary to confirm that the obtained MLR model allows to predict the concentration of the compound in new solutions, which is why the solution test group was used, since they are solutions apart from the model selection process.
In order to elucidate this issue, we have highlighted the keywords “low selectivity and cross-sensitivity sensors” in two sentences at:
Line 209, page 6; and, Line 226, page 6.
12) The same question occurs for the results presented in Figure 4.
In order to elucidate this point the text of Figures’ legend:
“Figure 4. - Quantification of sugars (glucose, fructose and sucrose) and total sugar content using a potentiometric sensor array and MLR models for the train (dotted line and solid marker) and test data (solid line and circle marker): A) standard solutions with KCl 1 mol L-1; B) standard solutions without KCl 1 mol L-1.”
Was changed to (line 390, page 13):
“Figure 4. - Linear relations between the predicted concentrations of sugars (glucose, fructose and sucrose) and total sugar content by the MLR models, using the sensors selected by the simulated annealing algorithm, and expected concentrations for the train (dotted line and solid marker) and test data (solid line and circle marker): A) standard solutions with KCl 1 mol L-1; B) standard solutions without KCl 1 mol L-1.”
13) Even though some differences are expected from the duplicated systems, I
was surprised with the results presented at Table 3. On average, 16 sensors
were the best fit for each sugar, however, only 4 to 5 sensors from ET1 were
also selected from ET2. Example of selected sensors for fructose (with KCl):
ET1 – 4, 8, 12, 14, 15 and 16; ET2 – 2, 3, 8, 14, 15, 16, 17, 18, 19 and 20.
Common (duplicated) sensors: 8, 14, 15 and 16.
A higher number of common sensors were expected when a duplicated system
is used (higher repeatability on sensors response / assembly). Could authors
comment on that?
The purpose of a duplicated system application was to allow the selection of the best sensor subset, emphasizing the selection of the smallest number of sensors, which allows compound concentration prediction using the MLR model. Common sensors in the MLR model is advantageous mainly in similar matrix samples were the most adequate low selectivity and cross-sensitivity sensors improves MLR prediction performance.
14) Based on the results described at Table 3, which sensor array should be
used for sugar quantification?
This question is not easy to answer because the sensors present in the duplicate system are practically all used in the various MLR models presented. I should also emphasize that the MLR models presented in this study are associated with the influence of the matrix of the solutions analyzed in this study, which are simple because they are synthetic solutions. But in the case of real samples the MLR models will be different from those obtained, with selection of other sensor subsets as the complexity of the sample matrix will influence the signals of the electronic tongue sensors. Thus, it is implied that the selection of a set of low selectivity and cross-sensitivity sensors for a given quantitative analytical study is only possible if the solution matrix has controlled variability. In this study, as referred, all the sensors present in the duplicate system were practically used due mainly to the difficulty in sensing sugars. This work allows us to infer that it is possible to quantify sugars if they are the majority compounds in the sample, such as, in honeys and juice drinks.
15) Have the authors evaluated system performance (sugar quantification)
using just the data collected with ET1 (or ET2) or both systems are needed?
Yes. The results were similar between for both of them. However, using a double system coupled with a metaheuristic algorithm for variable selection allows to improve robustness of multivariate model predictive performance. Using only ET1 or ET2, the prediction results for the solutions test group are not satisfactory.
Considering these last points, if the reviewer considers this information pertinent, I hope to have the opportunity to make the necessary changes.
Author Response File: Author Response.docx
Reviewer 3 Report
Dear Authors,
First of all, congratulations for this interesing manuscript. I have been pleased to review it. The work is innovative, clear, well written and high in quality. Nevertheless, in order to improve the manuscrit, please let me suggest some changes and aspects to be discussed. Let me write them point by point:
- First paragraph section 2.2. Standard solutions. Could you please provide more information on why do you select these 5 levels of sugar concentrations and not any others? Are they typical sugar concentrations in some specific food, fruit or juice?
Section 2.3. Please consider providing a picture of the sensor array. It would be very selfexplanatory.
Have you conducted any study to analyze the repeteability of the assays? Please clarify…
Additionally, have you considered studying other natural minority sugars such as pentose, xylose or arabinose? They are minor but also important. Please justify.
I have been impressed for the results of the MLR analyses and I am longing to know about future results on real samples.
Again, congratulations!!!
Best regards.
Author Response
Reviewer 3
Changes were carried out using the color blue.
First of all, congratulations for this interesting manuscript. I have been pleased to review it. The work is innovative, clear, well written and high in quality.
We kindly acknowledge the Reviewer’s positive comments. We thank all the efforts made through the correction process, since its suggestions were a major contribution to improve the present manuscript. New corrections were made throughout the revised Manuscript and we expect that the concerns of the reviewer have been fulfilled.
Nevertheless, in order to improve the manuscrit, please let me suggest some changes and aspects to be discussed. Let me write them point by point:
- First paragraph section 2.2. Standard solutions. Could you please provide more information on why do you select these 5 levels of sugar concentrations and not any others?
The sugars’ concentration levels were selected considering a previous work of sugars analysis by HPLC. To elucidate this point the phrase in line 130, page 3:
“The individual sugar concentrations were within the dynamic range of concentration of a HPLC method validated in a previous work [14].”
was changed to (line 144, page 4):
“The individual sugar concentrations were within the dynamic range of concentration of a HPLC method validated in a previous work for soft drinks analysis [14].”
Are they typical sugar concentrations in some specific food, fruit or juice?
Yes. The sucrose, fructose and glucose are typical sugars (main compounds) found in soft drinks, juice beverages, honeys and fermentative processes. They are matrix solutions in which we are interested in applying electronic tongue as an analytical quality control tool and also as a process monitoring tool.
Section 2.3. Please consider providing a picture of the sensor array. It would be very selfexplanatory.
A new figure (Figure 1) was introduced in the Revised Manuscript showing the sensor schematics of the electronic tongue. The figure was inserted in the end of the revised manuscript.
Figure 1 – Sensor schematics of the electronic tongue.
Have you conducted any study to analyze the repeteability of the assays? Please clarify...
The repeatability of sensors (data not included in the article) in the electronic tongue is always verified before starting a research work. To clarify this point, a new sentence has been added to page 8, line 288:
“Overall, the lipid polymeric membranes on the sensor arrays had satisfactory signal stability over time analysis (during 5 min of signal record), with relative standard deviations (%RSD) lower than 3.5%, and repeatability, with %RSD values between 0.5 and 12%.“
Additionally, have you considered studying other natural minority sugars such as pentose, xylose or arabinose? They are minor but also important. Please justify.
This work allows us to infer that it is possible to quantify sugars but since the electronic tongue is built with of low selectivity and cross-sensitivity sensors, it is implied that sugars analysis can be carried out if they are the majority compounds in the sample, such as, in honeys and juice drinks. To analyze minor compounds in a real sample would very difficult considering that the solutions’ matrix variability overlaps their information.
I have been impressed for the results of the MLR analyses and I am longing to know about future results on real samples.
Thank you for your positive comments.
As future work, we will apply this methodology to honey analysis. Considering that these sugars are the majority compounds, we hope to succeed.
Author Response File: Author Response.docx
Round 2
Reviewer 2 Report
First of all, I would like thank the authors for addressing all the issues previously pointed. However, there are still some issues that need better clarification probably because I was not very clear.
Regarding sensor reset I meant about sensor signal recovery after the cleaning step. Does the sensor response array return to its reference value after measuring the samples? Is there any signal drift?
Considering between inter-day analysis, the sensors are kept soaked in HCl solution? Do the authors do that for storage as well (longer periods with no measurements)?
In my previous comments (question 13) I inquired about the duplicated system, and the reduced number of identical sensors used to evaluate the data (this was already addressed by the authors). However, I missed one point, what is the influence of the duplicated sensors on the MRL model? (they are kind of redundant).
Nevertheless, authors are encouraged to further discuss the array used for the model (at least 16 sensors). Even though authors use 16 different physical sensors, several of them have the same membrane.
Considering their MRL model output, have the authors measured unknown samples (blind tests)? Authors are encouraged to do them, beacuse that would be very important to attest the effectiveness of their ET system in quantifying glucose, fructose and sucrose in ternary mixture solutions.
Lastly, Table 4 is much better now, however authors are encouraged to write column 4 (R2) values in a single line. Also, Figure numbers must be updated after the inclusion of the new Figure 1.
Author Response
Changes were carried out using the olive green color.
Comments and Suggestions for Authors
First of all, I would like thank the authors for addressing all the issues previously pointed. However, there are still some issues that need better clarification probably because I was not very clear.
We kindly acknowledge the Reviewer’s positive comments. We thank all the efforts made through the correction process, since its suggestions were a major contribution to improve the present manuscript. New corrections were made throughout the revised Manuscript and we expect that the concerns of the reviewer have been fulfilled.
Regarding sensor reset I meant about sensor signal recovery after the cleaning step. Does the sensor response array return to its reference value after measuring the samples? Is there any signal drift?
Yes, there is a signal drift but not significant if all the analysis were carried out in the same day.
To minimize the risk of high drifts the total time employed for the all samples analysis was less than 8 hours (in the same day). Although this will not eliminate drift issues it can reduce their impact. Indeed, assays concerning signal’s time stability during a 3-h continuous sample analysis showed that the maximum percentage signal variation, between the initial signal (after signal stabilization) and that recorded after the 3-h, was lower than 4.8 %, corresponding to a potential variation of 4 mV.
Also, daily drift is also minimized by using signal profile from a sensors array coupled with multivariate statistical techniques that allow its compensation, since signal drift will be a common point in the signals of the various sensors (practically constant variation), considering that the variation between the signals of the selected sensor profile is the feature used in the MLR model for the concentration prediction of a compound.
However, if the analyses are done in different days or months, we should use a standard solution with constant concentration to solve signal drifts, by subtracting the samples E-tongue’s signal profile by the respective signal profile recorded for the standard solution, obtained in the same day.
In a previous work [Ref1] of olive oil analysis, samples were analyzed during one year (0, 3, 6, 9 and 12 months) and, in order to control the potentiometric signal drifts of the E-tongue sensors, a standard solution (gallic acid, 1×10-3mol/L) was analyzed before and after each olive oil measurement series. Signal drifts were solved by subtracting the signal profile recorded by the E-tongue device during the analysis of each olive oil sample by the average signal profile recorded for the gallic acid standard solution.
[Ref1] N. Rodrigues, L.G. Dias, A.C.A. Veloso, J.A. Pereira, A.M. Peres, Evaluation of extra-virgin olive oils shelf life using an electronic tongue-chemometric approach, European Food Research and Technology 243 (2017) 597–607; DOI: 10.1007/s00217-016-2773-2.
A new phrase was introduced in the revised manuscript (page 8, line 291) to elucidate about the sensor drift:
“Moreover, assays concerning signal’s time stability during a 3-h continuous sample analysis showed that the maximum percentage signal variation, between the initial signal (after the 5 minutes of signal stabilization) and that recorded after 3 hours, was lower than 4.8 %, corresponding to a potential variation of 4 mV. These overall results confirmed that the signal drift was not significant in intra-day analyzes.”
Considering between inter-day analysis, the sensors are kept soaked in HCl solution? Do the authors do that for storage as well (longer periods with no measurements)?
Yes. Between inter-days and even for long periods without any analysis, the E-tongue is kept immersed in HCl solution.
In my previous comments (question 13) I inquired about the duplicated system, and the reduced number of identical sensors used to evaluate the data (this was already addressed by the authors).
However, I missed one point, what is the influence of the duplicated sensors on the MRL model? (they are kind of redundant).
This question is pertinent and so it was included new text to clarify this issue.
In fact, duplicate sensors should give the same signals towards the same sample. But, duplicates may have slightly different signals behaviors towards the same sample. In this case, duplicated sensors are no redundant since each one is considered as an independent sensor in the MLR model. So, in several models duplicates where included (meaning that signals may contain relevant or complementary information regarding the analyzed samples) and for other sensors only one of the duplicates was used (possibly meaning that those duplicates were in fact identical – “real” replicas - and gave very similar results, that would be highly correlated and so, could not be considered as independent variables by the model).
In Page 8, line 301 (old manuscript), it was referred that “These differences can be attributed to slight differences in the polymeric membranes formation due to the drop-by-drop technique used.”. This aspect is important to explain that some of the slight differences observed between signals recorded by a sensor and its replica may be attributed to slight variations that may occur, due to the drop-by-drop technique applied, in the membrane’s thickness and therefore in their physical properties.
So, the referred phrase was changed to (page 9, line 372):
“These differences can be attributed to slight variations in the lipid polymeric membranes formation, which may occur by using the drop-by drop technique, due to variations in the membrane mixture solution viscosity and its solvent’s evaporation temperature, contributing to changes in the physical properties of the membrane (e.g., membrane thickness, transparency degree and porosity).”
Nevertheless, authors are encouraged to further discuss the array used for the model (at least 16 sensors). Even though authors use 16 different physical sensors, several of them have the same membrane.
In which concerns the model establishment, sensors are included by numerical reason. In fact, since these sensors have cross-sensitivity and are not selective to any particular compound, in principle, each sensor give a different overall fingerprint of the beverage matrix, and so, may contribute to beverage identification having similar probability to be included in the model. However, in some cases the inclusion of a particular sensor could be tentatively explained chemically, considering the nature (e.g., functional groups present) of the plasticizer/additive. For example, in the case of sugars, their detection by the E-tongue could be attributed to the establishment of possible electrostatic interactions between, hydroxyl groups of sugars molecules and the functional groups (e.g., phosphate) in the membrane surface, by means of hydrogen bonds. Unfortunately, has also recognized by some Authors (e.g., Toyota et al., 2011) the full mechanism behind this potentiometric behavior of lipid/polymer membranes is not well known.
Related to this point, two new phrases were added, as well 4 new references:
Page 11, line 363:
“All MLR models had duplicate sensors included, which means that their signals may contain relevant or complementary information regarding the analyzed samples. These sensors are considered as independent variables or false replicas by the multivariate model, since they were not highly correlated as a true replicate should be.”
Page 8, line 295:
“The E-tongue sensitivity to the three sugars may be related to the nature (e.g., functional groups) of the plasticizer/additive present in each polymeric membrane of the sensor array, once the full mechanism behind this potentiometric behavior of lipid/polymer membranes is not well known [39]). The lipid polymeric membranes used contain hydrophobic and hydrophilic groups allowing the interaction with electrolytes and nonelectrolytes chemical compounds via hydrophobic or electrostatic interactions [39,40,41], being also possible to arise hydrogen bonds [42]. For instance, the sugars detection could be attributed to the establishment of possible electrostatic interactions between, hydroxyl groups of sugars molecules and the functional groups (e.g., between carboxyl or phosphate groups) by means of hydrogen bonds.”
[39] K. Toyota, H. Cui, K. Abe, M. Habara, K. Toko, H. Ikeazaki, Sweetness sensor with lipid/polymer membranes: sweet-responsive substances, Sensor. Mater. 23 (2011) 465–474.
[40] Y. Kobayashi, M. Habara, H. Ikezazki, R. Chen, Y. Naito, K. Toko, Advanced taste sensors based on artificial lipids with global selectivity to basic taste qualities and high correlation to sensory scores, Sensors 10 (2010) 3411–3443.
[41] M. Yasuura, H. Okazaki, Y. Tahara, H. Ikezaki, K. Toko, Development of sweetness sensor with selectivity to negatively charged high-potency sweeteners, Sensor. Actuat. B 201 (2014) 329–335.
[42] K. Toyota, H. Cui, K. Abe, M. Habara, K. Toko, H. Ikeazaki, Sweetness sensor with lipid/polymer membranes: response to various sugars, Sensor. Mater. 23 (2011) 475–482.
Considering their MRL model output, have the authors measured unknown samples (blind tests)? Authors are encouraged to do them, because that would be very important to attest the effectiveness of their ET system in quantifying glucose, fructose and sucrose in ternary mixture solutions.
Yes. These unknown samples were represented by samples from the test group, which were not used to obtain the MLR model. Moreover, in order to reduce solution preparation errors and to have no dependence between the concentrations of the three analyzed sugars, solutions were prepared independently following an orthogonal experimental design (a fractional factorial design), using measured mass mixtures of sugars rather than dilutions of stock solutions.
In order to elucidate this issue, the phrase in page 6, line 219: “The standard solutions were split into train group, for MLR model building, and test group (external evaluation set), for testing externally the predictive performance of the obtained model.”,
was changed to:
“The standard solutions were split into train group, for MLR model building, and test group (independent external evaluation set), considered as unknown samples for testing externally the predictive performance of the obtained model.”
Lastly, Table 4 is much better now, however authors are encouraged to write column 4 (R2) values in a single line. Also, Figure numbers must be updated after the inclusion of the new Figure 1.
In the table 4, I rearrange the variables identification format into two lines in order to fit the table within the page width and keep all values in a line
Round 3
Reviewer 2 Report
The revised manuscript is very well written and all the issues were clarified by the authors. Once again, I would like to thank the authors for the fruitful discussion and clarifications provided.
In my opinion the manuscript, in its current form, can be accepted for publication in Chemosensors.
