Voltammetric Fingerprinting and Chemometrics: A Rapid and Robust Platform for Ground Clove Bud Authentication and Adulteration Detection

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This manuscript presents a voltammetric fingerprinting approach using an MWCNT-modified electrode combined with chemometric analysis (PCA, PLS-DA, HCA, and PLSR) for geographical authentication and adulteration detection in ground clove bud. The topic is relevant to food authentication, and the experimental design is generally clear. However, in its current form, the manuscript requires major revision before it can be considered for publication in Chemosensors.

From a chemometric perspective, the main weakness of the manuscript is the lack of interpretation of the latent variables. For geographical discrimination, the authors report that PC1 explains approximately 77% of the variance and contributes strongly to group separation. However, no loading analysis is presented. If PC1 carries the primary discriminatory information, its loadings must be evaluated to understand which potential regions or electrochemical signals are responsible for the observed separation. Without examining PC1 loadings, the discrimination remains purely statistical and lacks chemical interpretation. The same issue applies to PLS-DA. After preprocessing, Factor 1 appears to clearly separate the three regions, while Factor 2 contains additional information distinguishing East Java from the other regions. This is an important analytical finding, yet no discussion is provided regarding which voltammetric features contribute to these factors. A deeper chemical interpretation based on loadings or regression coefficients is necessary to support the proposed authentication mechanism.

There are also inconsistencies in figure referencing. On page 10, the manuscript refers to “Figure 3d,” which does not exist. This must be corrected for clarity and consistency.

In addition, the dendrograms shown in Figures 6 and 7 lack readable sample names. The labels are not visible or are too small to interpret properly, which prevents proper evaluation of clustering behavior. Clear and legible sample identification is essential for validating the conclusions drawn from HCA.

Throughout the manuscript, the authors frequently refer to improvements “after preprocessing” without consistently specifying which preprocessing method was applied. Given that multiple strategies were tested (mean normalization, maximum normalization, autoscaling, combinations), it is essential to explicitly state which preprocessing method corresponds to each reported result. General statements such as “after preprocessing” are insufficient and reduce reproducibility.

Finally, although PLSR models show acceptable calibration and cross-validation performance, validation remains internal. No independent external samples, blind testing, or real-market adulterated samples were included. This limits the robustness claims and practical applicability of the proposed platform.

In summary, while the study presents a technically sound application of voltammetric fingerprinting to clove bud authentication, significant improvements are required. In particular, deeper chemometric interpretation (loadings analysis), clearer reporting of preprocessing strategies, correction of figure inconsistencies, improved visualization of dendrogram labels, and strengthened validation are necessary. I therefore recommend major revision before further consideration.

Author Response

Reviewer 1

This manuscript presents a voltammetric fingerprinting approach using an MWCNT-modified electrode combined with chemometric analysis (PCA, PLS-DA, HCA, and PLSR) for geographical authentication and adulteration detection in ground clove bud. The topic is relevant to food authentication, and the experimental design is generally clear. However, in its current form, the manuscript requires major revision before it can be considered for publication in Chemosensors.

1. From a chemometric perspective, the main weakness of the manuscript is the lack of interpretation of the latent variables. For geographical discrimination, the authors report that PC1 explains approximately 77% of the variance and contributes strongly to group separation. However, no loading analysis is presented. If PC1 carries the primary discriminatory information, its loadings must be evaluated to understand which potential regions or electrochemical signals are responsible for the observed separation. Without examining PC1 loadings, the discrimination remains purely statistical and lacks chemical interpretation. The same issue applies to PLS-DA. After preprocessing, Factor 1 appears to clearly separate the three regions, while Factor 2 contains additional information distinguishing East Java from the other regions. This is an important analytical finding, yet no discussion is provided regarding which voltammetric features contribute to these factors. A deeper chemical interpretation based on loadings or regression coefficients is necessary to support the proposed authentication mechanism.

Answer:

We thank the reviewer for this insightful and constructive comment. We agree that the interpretation of the latent variables is important to provide a clearer chemometric and chemical understanding of the discrimination mechanism. In response to this suggestion, we have examined the loading profiles of the PCA and PLS-DA models to identify the voltammetric variables contributing to the observed sample separation.

Additional discussion has been incorporated in the revised manuscript to interpret the latent variables in terms of the most influential potential regions of the voltammetric fingerprints. Specifically, the loading analysis indicates that PC1, which explains the largest proportion of the variance, is mainly associated with voltammetric variables located in particular potential regions that contribute strongly to the separation observed in the PCA score plot. PC2 provides additional discrimination among the samples, particularly separating North Maluku samples from those from South Sulawesi and East Java.

A similar analysis was performed for the PLS-DA model. The X-loading profiles of Factor 1 and Factor 2 were examined to identify the voltammetric variables responsible for the discrimination among geographical origins. The interpretation highlights that Factor 1 captures the main variance related to electrochemical signals of the clove extracts, while Factor 2 provides additional discriminatory information among the regions.

To support this interpretation, the loading plots of PC1 and PC2 (PCA) and the X-loading plots of Factor 1 and Factor 2 (PLS-DA) have been included in the Supplementary Material (Figures S2 and S3). These plots illustrate the contributions of voltammetric variables (current responses at different potentials) to the latent variables underlying sample discrimination.

Furthermore, the revised manuscript now includes a brief discussion linking the prominent voltammetric regions to electrochemical signals likely associated with the redox processes of electroactive compounds in clove extracts, such as phenolic constituents. These additions provide a clearer interpretation of the latent variables and strengthen the chemical basis of the proposed authentication approach.

The corresponding clarifications and discussions have been added and highlighted in the revised manuscript page 8 line 301-319.

 

2. There are also inconsistencies in figure referencing. On page 10, the manuscript refers to “Figure 3d,” which does not exist. This must be corrected for clarity and consistency.

Answer:

We thank the reviewer for pointing this out. The reference to “Figure 3d” on page 10 was incorrect and should refer to Figure 4b. This has been corrected in the revised manuscript page 10, and the change has been highlighted.

 

3. In addition, the dendrograms shown in Figures 6 and 7 lack readable sample names. The labels are not visible or are too small to interpret properly, which prevents proper evaluation of clustering behavior. Clear and legible sample identification is essential for validating the conclusions drawn from HCA.

Answer:

We thank the reviewer for this important observation. The dendrograms in Figures 6 and 7 have been revised to improve the readability of the sample labels. The sample names have been reformatted and enlarged to ensure clear and legible identification, enabling proper evaluation of the clustering patterns. The updated figures have been included in the revised manuscript page 13 and 14, respectively.

 

4. Throughout the manuscript, the authors frequently refer to improvements “after preprocessing” without consistently specifying which preprocessing method was applied. Given that multiple strategies were tested (mean normalization, maximum normalization, autoscaling, combinations), it is essential to explicitly state which preprocessing method corresponds to each reported result. General statements such as “after preprocessing” are insufficient and reduce reproducibility.

Answer:

We thank the reviewer for this important comment regarding reproducibility. The preprocessing strategies applied in each chemometric analysis are in fact explicitly specified in the manuscript. For clarity: PCA for geographical discrimination was performed using backward-scan current data after maximum normalization combined with autoscaling (explained in page 8 line 298-300); PLS-DA was built using mean normalization applied to forward-scan data (explained in page 9 line 329-331); for adulteration discrimination, the optimal preprocessing was maximum normalization combined with autoscaling (explained in page 11 line 402-405); and for PLSR modelling, mean normalization with autoscaling (stem adulteration) and autoscaling of forward-scan data (soil adulteration) were used (explained in page 15 line 488-491). These preprocessing conditions are described in the respective sections where the corresponding models are presented. Nevertheless, we carefully reviewed the manuscript to ensure that the applied preprocessing method is clearly indicated whenever results “after preprocessing” are discussed, thereby improving clarity and reproducibility.

 

 

 

 

5. Finally, although PLSR models show acceptable calibration and cross-validation performance, validation remains internal. No independent external samples, blind testing, or real-market adulterated samples were included. This limits the robustness claims and practical applicability of the proposed platform.

Answer:

We thank the reviewer for this valuable suggestion and appreciate the comment on including independent external samples or conducting blind testing to further strengthen the robustness of the proposed model. In the present study, model performance was evaluated using calibration and internal cross-validation, which are commonly applied approaches in exploratory chemometric studies. We agree that including independent external samples and real-world adulterated samples would provide additional confirmation of the model’s predictive capability and practical applicability. However, this aspect falls beyond the scope of the current study and will be considered in future work to further validate and extend the proposed analytical platform.

In summary, while the study presents a technically sound application of voltammetric fingerprinting to clove bud authentication, significant improvements are required. In particular, deeper chemometric interpretation (loadings analysis), clearer reporting of preprocessing strategies, correction of figure inconsistencies, improved visualization of dendrogram labels, and strengthened validation are necessary. I therefore recommend major revision before further consideration.

 

 

 

                                                                   

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The paper, titled " Voltammetric Fingerprinting and Chemometrics: A Rapid and Robust Platform for Ground Clove Bud Authentication and Adulteration Detection," is a very well-done paper. The Authors use electrodes appropriately modified with carbon nanotubes, which prove to be effective. They also perform a thorough and comprehensive characterization of the nanotubes used. The chemometric methods chosen are appropriate and well-developed. The work is very interesting for its potential applications in the fight against food fraud, which is unfortunately widespread in the case of high-quality spices. I have only a few comments and suggestions, listed below.

Abstract. I don’t agree the proposed method is “nondestructive”, because it requires to solubilize samples.

Line 129: Please specify % w/w or v/v and the other solvent (although the reader may assume it is deionized water)

Line 142 and 144: put “-1” in superscript

Line 154: ERRATA: “GCE modified” CORRIGE: “GCE was modified”

Line 290: I can’t find S1

Paragraph 2.4 : Was deaeration performed prior to analysis? If so, please specify.

Figures 3a and 3b. I think the comments regarding the Figure 3a (lines 292-294 of the text) are pessimistic. In fact, Figure 3a shows the three centroids quite far apart. Rather than " insufficient" I would use "improvable." In fact, my comments are confirmed by the Authors at line 386 (“Although PLS-DA is a supervised classification technique, its performance in this study does not differ markedly from PCA). Analogously, referring to Figure 3b, I advise to change the word “clear” at line 298 and use “satisfying” instead; in my opinion, in fact, the separation IS clear, but it can be further improved. I suggest instead not to use different scales for the two axes in Figures 3a and 3b, as this could be misleading; in fact, it distorts the actual distances between the points. I would therefore recommend using the same scale (-1200 to +1200) on both axes.

Lines 308 and 309: recalling my comments to Figure 3, change “distinct” to “well distinct” and “incomplete” into “unsatisfactory”

Figure 4. Similar comment to the one referred to in Figure 3, regarding the axis scales.

Line 325: change: “unseen data” to: “unknown samples”

Line 330: ERRATA: “Figure 3d” CORRIGE: “Figure 4b”

Table 1: Remove the 0 factor. Use dot “.” not comma “,” for decimals. Don't use too many digits for variance values; I find rounding them to two decimal places is sufficient. This isn't a mistake, but reducing the displayed digits makes the table more readable. For example, change "98.2165451" to "98.22".

Line 352: I recommend changing “chemometric value” to “chemometric information”

Figure 5. Similar comment to the one referred to in Figures 3 and 4, regarding the axis scales.

Line 402: please specify which type of distance was used for HCA (Euclidean? Mahalanobis? Others?). Did the Authors use the original variables or scores to calculate distances? Please, specify.

Line 438: I can’t find S2

Line 443: I can’t find S3 and S4

Line 451: ERRATA: “The calibration plots” CORRIGE: “The control plots Experimental vs. Calculated” or “The control plots Experimental vs. Predicted”

Figures 8 and 9: I suggest not to use different scales for the two axes, because this prevents from evidencing that the target line is that with null intercept and unitary slope. I would therefore recommend using the same scale on both axes.

Table 2. A null error does not make sense. I guess that the null values in the last columns “bias” correspond to values below ten to the minus 4. Moreover, the numbers have non-uniform formats; for instance 0.9436 shows four digits, while 0.0146 shows three digits. I suggest to use three digits for all numbers: for instance, 0.9436 becomes 0.944; for numbers in the column “bias”, format them by the scientific notation, using three digits for the mantissa (e.g. 1.234·10-4, with “-4” in superscript).

Author Response

Reviewer 2

The paper, titled " Voltammetric Fingerprinting and Chemometrics: A Rapid and Robust Platform for Ground Clove Bud Authentication and Adulteration Detection," is a very well-done paper. The Authors use electrodes appropriately modified with carbon nanotubes, which prove to be effective. They also perform a thorough and comprehensive characterization of the nanotubes used. The chemometric methods chosen are appropriate and well-developed. The work is very interesting for its potential applications in the fight against food fraud, which is unfortunately widespread in the case of high-quality spices. I have only a few comments and suggestions, listed below.

1. I don’t agree the proposed method is “nondestructive”, because it requires to solubilize samples.

Answer:

We thank the reviewer for this observation. We agree that the term “nondestructive” is not fully appropriate since the analysis requires extraction of the sample in a solvent. Therefore, the term has been removed and replaced with “practical” in the Abstract. This modification has been implemented and highlighted in the revised manuscript page 1 line 35.

 

2. Line 129: Please specify % w/w or v/v and the other solvent (although the reader may assume it is deionized water)

Answer:

We thank the reviewer for this comment. The solvent composition has been clarified in the revised manuscript. Specifically, the ethanol concentration is now expressed as 50% (v/v), and the other solvent has been specified as deionized water. The revised text has been updated and highlighted in the manuscript page 3 line 129.

 

3. Line 142 and 144: put “-1” in superscript

Answer:

We appreciate the reviewer’s careful observation. The exponent “−1” has been corrected to superscript format in the revised manuscript. The corresponding changes have been highlighted in page 4 line 143-144, for clarity.

 

4. Line 154: ERRATA: “GCE modified” CORRIGE: “GCE was modified”

Answer:

We thank the reviewer for this correction. The phrase “GCE modified” has been revised to “GCE was modified” to ensure grammatical accuracy. The change has been implemented and highlighted in the revised manuscript page 4 line 155.

 

5. Line 290: I can’t find S1

Answer:

We thank the reviewer for pointing this out and apologize for the oversight. The Supplementary Material containing Figure S1 was inadvertently omitted during the submission process. The Supplementary Material has now been uploaded and is available with the revised manuscript. The corresponding reference in the manuscript has also been checked for consistency.

 

6. Paragraph 2.4 : Was deaeration performed prior to analysis? If so, please specify.

Answer:

We appreciate the reviewer’s comment. No deaeration was performed prior to the voltammetric measurements. This information has now been added and highlighted in Section 2.4 of the revised manuscript page 4 line 158-159.

 

7. Figures 3a and 3b. I think the comments regarding the Figure 3a (lines 292-294 of the text) are pessimistic. In fact, Figure 3a shows the three centroids quite far apart. Rather than " insufficient" I would use "improvable." In fact, my comments are confirmed by the Authors at line 386 (“Although PLS-DA is a supervised classification technique, its performance in this study does not differ markedly from PCA). Analogously, referring to Figure 3b, I advise to change the word “clear” at line 298 and use “satisfying” instead; in my opinion, in fact, the separation IS clear, but it can be further improved. I suggest instead not to use different scales for the two axes in Figures 3a and 3b, as this could be misleading; in fact, it distorts the actual distances between the points. I would therefore recommend using the same scale (-1200 to +1200) on both axes.

Answer:

We thank the reviewer for this insightful comment and for the constructive suggestions regarding the interpretation of Figures 3a and 3b. We agree that the separation observed in Figure 3a indicates that the centroids of the three groups are reasonably separated; therefore, the term “insufficient” has been replaced with “improvable” to provide a more balanced interpretation. This modification better reflects that, while some discrimination is already present, further improvement is still possible.

In addition, following the reviewer’s recommendation, the word “clear” to describe Figure 3b has been revised to “satisfying” to present a more nuanced description of the clustering result. These revisions are also consistent with our discussion, in which we note that PLS-DA performance does not differ markedly from PCA. All corresponding changes have been implemented and highlighted in the revised manuscript page 8 and page 9.

We also thank the reviewer for this important observation regarding the axis scaling. Following this suggestion, the axes in Figures 3a and 3b have been adjusted to use the same scale in order to avoid potential distortion in the visual representation of distances between points. The figures have been revised accordingly in the updated manuscript page 9.

 

8. Lines 308 and 309: recalling my comments to Figure 3, change “distinct” to “well distinct” and “incomplete” into “unsatisfactory”

Answer:

We thank the reviewer for this suggestion. Following the reviewer’s comment regarding Figure 3, the term “distinct” has been revised to “well distinct,” and “incomplete” has been replaced with “unsatisfactory” to better reflect the intended meaning. These changes have been implemented and highlighted in the revised manuscript page 9.

 

9. Figure 4. Similar comment to the one referred to in Figure 3, regarding the axis scales.

Answer:

We appreciate the reviewer’s comment regarding the axis scales in Figure 4. In accordance with this suggestion, the axes of Figures 4a and 4b have been modified to use identical scaling to avoid potential distortion in the graphical representation of the distances between samples. The revised figures have been updated in the manuscript page 10.

 

 

 

10. Line 325: change: “unseen data” to: “unknown samples”

Answer:

We thank the reviewer for this suggestion. The term “unseen data” has been revised to “unknown samples” in the revised manuscript page 10 line 360. The change has been implemented and highlighted for clarity.

 

11. Line 330: ERRATA: “Figure 3d” CORRIGE: “Figure 4b”

Answer:

We thank the reviewer for identifying this error. The reference to “Figure 3d” was incorrect and has been corrected to Figure 4b in the revised manuscript page 10. The change has been highlighted for clarity.

 

12. Table 1: Remove the 0 factor. Use dot “.” not comma “,” for decimals. Don't use too many digits for variance values; I find rounding them to two decimal places is sufficient. This isn't a mistake, but reducing the displayed digits makes the table more readable. For example, change "98.2165451" to "98.22".

Answer:

We appreciate the reviewer’s suggestion to improve the readability of Table 1. The unnecessary factor “0” has been removed, decimal commas have been replaced with decimal points, and the variance values have been rounded to two decimal places as recommended (e.g., “98.2165451” revised to “98.22”). These modifications have been implemented in the revised manuscript page 10 to enhance the clarity and presentation of the table.

 

13. Line 352: I recommend changing “chemometric value” to “chemometric information”

Answer:

We thank the reviewer for this helpful suggestion. The term “chemometric value” has been revised to “chemometric information” to improve clarity and accuracy. The change has been implemented and highlighted in the revised manuscript page 11 line 387.

 

14. Figure 5. Similar comment to the one referred to in Figures 3 and 4, regarding the axis scales.

Answer:

We thank the reviewer for this helpful comment. Following the reviewer’s suggestion and consistent with the revisions to Figures 3 and 4, the axis scales in Figure 5 have been revised to improve clarity and ensure consistency across the axes. The updated figure has been included in the revised manuscript page 12.

 

 

 

15. Line 402: please specify which type of distance was used for HCA (Euclidean? Mahalanobis? Others?). Did the Authors use the original variables or scores to calculate distances? Please, specify.

Answer:

We thank the reviewer for this important comment. The hierarchical cluster analysis (HCA) parameters have now been specified in the revised manuscript. For the stem adulteration analysis, HCA was performed using the current values at oxidation potentials ranging from −0.4 to +1.0 V vs. Ag/AgCl, after mean normalization preprocessing. The distances were calculated using the normalized Manhattan distance, and clustering was carried out using the Ward linkage method. For the soil adulteration analysis, HCA was conducted using all recorded current values after maximum normalization preprocessing. In this case, the normalized Euclidean distance was used together with the Ward linkage method. These details have been added to the revised manuscript page 12 line 441-442 and page 13 line 456-458 for clarity.

 

16. Line 438: I can’t find S2

Answer:

We thank the reviewer for pointing this out and apologize for the oversight. The Supplementary Material containing Figure S2 was inadvertently omitted during the initial submission. It has now been uploaded and is available with the revised manuscript.

 

17. Line 443: I can’t find S3 and S4

Answer:

We appreciate the reviewer for noticing this issue. The Supplementary Materials corresponding to Figures S3 and S4 were unintentionally not included in the original submission. These files have now been provided and are included with the revised manuscript.

 

18. Line 451: ERRATA: “The calibration plots” CORRIGE: “The control plots Experimental vs. Calculated” or “The control plots Experimental vs. Predicted”

Answer:

We thank the reviewer for this helpful suggestion. To improve clarity and consistency with the figure labels, the phrase “The calibration plots” has been revised to “The Predicted vs. Reference control plots.” This modification has been implemented and highlighted in the revised manuscript page 15.

 

19. Figures 8 and 9: I suggest not to use different scales for the two axes, because this prevents from evidencing that the target line is that with null intercept and unitary slope. I would therefore recommend using the same scale on both axes.

Answer:

We thank the reviewer for this helpful suggestion. The axes in Figures 8 and 9 have been adjusted to use the same scale for both the X and Y axes. This modification allows a clearer visualization of the agreement between predicted and reference values and facilitates the comparison with the ideal target line (slope = 1 and intercept = 0). The figures have been updated accordingly in the revised manuscript page 15.

 

 

20. Table 2. A null error does not make sense. I guess that the null values in the last columns “bias” correspond to values below ten to the minus 4. Moreover, the numbers have non-uniform formats; for instance 0.9436 shows four digits, while 0.0146 shows three digits. I suggest to use three digits for all numbers: for instance, 0.9436 becomes 0.944; for numbers in the column “bias”, format them by the scientific notation, using three digits for the mantissa (e.g. 1.234·10-4, with “-4” in superscript).

Answer:

We thank the reviewer for this careful observation. The bias values reported as “0” correspond to extremely small values that were rounded by the software. To improve clarity and consistency, the numerical formatting in Table 2 has been revised. All values are now reported to 3 decimal places, and bias values are expressed in scientific notation when appropriate. These modifications have been implemented and highlighted in the revised manuscript page 16.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Accept in present form