Review Reports

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

In this review, the authors highlight the application, technical challenges, global cohort initiatives, and future directions of metabolomics. The authors bridge metabolomic research with precision medicine and population health, addressing a critical gap in understanding dynamic physiological homeostasis beyond disease absence. However, there are some issues that the authors should pay attention to, and the detailed comments are listed as follows:

1. It is recommended to adhere to scientific guidelines, such as the PRISMA, to ensure a standardized approach for including and excluding relevant literature in the manuscript.
2. The current description of limitations is too generic and superficial. The authors should provide a candid and specific discussion of the study’s limitations. The manuscript individually addresses determinants (age, sex, BMI, etc.) but inadequately explores their synergistic or antagonistic interactions.
3. Typing and labelling errors should be corrected.

Author Response

Comments 1:

It is recommended to adhere to scientific guidelines, such as the PRISMA, to ensure a standardized approach for including and excluding relevant literature in the manuscript.

Response 1:

(Please find uploaded Supplementary Table 1, "Representative molecular studies of human health", which we have put in great effort and gives an extensive summary of the studies and databases described in the manuscript.)

We thank the Reviewer 1 for this valuable suggestion. As this manuscript is a narrative and conceptual review without a dedicated Methods section and not a formal systematic review, we now explicitly describe a structured search approach informed by PRISMA principles to enhance transparency in study selection. In the Supplementary Table, we state:

"All of the studies was identified through a structured and transparent search strategy to minimize selection bias. Relevant studies were retrieved from PubMed, Web of Science, and Google Scholar using combinations of keywords including “metabolomics,” “metabolic profiling,” “health,” “healthy baseline,” “physiological homeostasis,” and “population cohort.” Priority was given to large-scale human cohort studies, representative methodological peer-reviewed papers, major talks/presentations personally given by the authors, and recent reviews that address metabolic variation in apparently healthy populations. This approach was informed by the general principles of the PRISMA guidelines, while allowing flexibility appropriate for a conceptual and integrative review.."

 

Comments 2:

The current description of limitations is too generic and superficial. The authors should provide a candid and specific discussion of the study’s limitations. The manuscript individually addresses determinants (age, sex, BMI, etc.) but inadequately explores their synergistic or antagonistic interactions.

Response 2:

Thank you for kindly pointing this out.

(i) While one has to fully respect the description and discussions written in each peer-reviewed publication, we have tried to provide analyses of each study's limitations, etc candidly. Please find these in Suppl Table 1, where we provide columns listing "Cohort samples", "Objectives", "Key publications" "Constraints", "Key Advantages" and "Methodology".

(ii) Furthermore, with regards to "exploring synergistic or antagonistic interactions", we have discussed this generally (with other limitations) in the main text - Please see Page 12, Lines 493-511:

"Finally, beyond technical and design-related challenges, defining metabolic health baselines is constrained by fundamental conceptual limitations. Traditional approaches often assume a static notion of “normal” metabolism, whereas metabolic homeostasis is inherently dynamic and continuously shaped by age, lifestyle, environmental exposures, and physiological adaptation. Consequently, a single static baseline may fail to capture meaningful variations in metabolic health across time and contexts.

Furthermore, metabolic profiles are highly context-dependent, reflecting the integrated effects of multiple interacting determinants rather than isolated factors. This context dependency complicates efforts to define universal reference ranges applicable across life stages and populations. Defining “healthy” metabolism is particularly challenging across the life course, as metabolic requirements and normative ranges differ substantially between childhood, adulthood, aging, and transitional states such as pregnancy or menopause.

Collectively, these limitations highlight that defining metabolic health baselines is constrained not only by data availability but also by how measurements are generated, standardized, and interpreted over time, reinforcing the need for dynamic, stratified, and context-aware reference frameworks."

 

 

 

Comments 3:

Typing and labelling errors should be corrected.

Response 3:

Thank you for kindly pointing this out and we deeply apologise for not proof-reading more carefully. We have extensively made changes throughout the manuscript.

Reviewer 2 Report

Comments and Suggestions for Authors

His manuscript addresses an important and timely topic—the definition of a metabolic baseline of health using metabolomics—and cites a broad range of studies. However, despite its ambition, the review falls short of the standards required for publication in its current form.

First, although the manuscript is presented as a narrative review, its coverage appears incomplete and selective, suggesting a biased rather than comprehensive overview of the field. I don't have any information about how the literature was identified, which databases were searched, what search terms were used, or how studies were selected or excluded. Given the breadth of the topic, this lack of transparency makes it difficult to assess the representativeness of the cited literature. At a minimum, a scoping review–level description of the search logic would be required; alternatively, the manuscript should be formally reframed as a scoping review with an explicit methodology.

Second, throughout the manuscript, the review primarily consists of a sequential listing of results from previous studies and extensive cohort studies without sufficient critical synthesis. Individual findings are described, but they are rarely interpreted in a way that yields deeper conceptual insight. Key questions—such as what fundamentally limits the definition of a metabolic health baseline, which conclusions are robust across platforms and cohorts, and which findings remain uncertain or contradictory—are not adequately addressed. As a result, the review reads more as an annotated bibliography than as a critical evaluation of the field.

Third, several core methodological issues are repeatedly acknowledged but not meaningfully analyzed. The manuscript emphasizes the need for large-scale and longitudinal data, yet does not critically review how such data can be generated and integrated in practice. Essential topics—such as long-term quality control, reproducible quantification across years, harmonization of data from different analytical platforms, and the separation of biological variability from measurement noise—are mentioned only in passing. Without addressing these issues from both measurement and data-analytical perspectives, the call for “larger datasets” remains generic and non-actionable.

Fourth, the discussion of analytical platforms is oversimplified. Mass spectrometry is effectively limited to LC–MS and GC–MS, while other established approaches, such as CE–MS, which are particularly relevant for highly polar and ionic metabolites central to core metabolism, are entirely omitted. This contributes to an artificial NMR-versus-MS dichotomy and further limits the methodological completeness of the review.

Fifth, although artificial intelligence and machine learning are repeatedly highlighted as key future enablers, no concrete examples are provided of AIーbased metabolomics advancing to routine clinical or real-world implementation. The challenges specific to metabolomic data—such as confounding by lifestyle, diet, and short-term physiological variation; overfitting in cross-sectional designs; and translation to low-cost, deployable assays—are not critically examined. Consequently, the AI discussion remains aspirational and indistinguishable from that found in many existing reviews.

Finally, the manuscript's conceptual contributionーis limited. Core ideas such as the dynamic nature of health, the role of metabolomics as a functional readout, and the importance of large cohorts and AI integration have been extensively discussed in prior literature. The manuscript does not clearly articulate how its perspective differs from or advances beyond existing reviews, nor does it offer a novel framework or set of principles that would justify publication as a standalone synthesis.

While the English language itself is generally clear and does not require extensive proofreading, the fundamental issues of scope, structure, methodological depth, and conceptual novelty cannot be resolved solely through revision. For these reasons, rejection is the most appropriate editorial decision.

Author Response

We are most grateful for the detailed comments by Reviewer 2, and have given our best efforts to address all points. 

 

Comments 1:

First, although the manuscript is presented as a narrative review, its coverage appears incomplete and selective, suggesting a biased rather than comprehensive overview of the field. I don't have any information about how the literature was identified, which databases were searched, what search terms were used, or how studies were selected or excluded. Given the breadth of the topic, this lack of transparency makes it difficult to assess the representativeness of the cited literature. At a minimum, a scoping review–level description of the search logic would be required; alternatively, the manuscript should be formally reframed as a scoping review with an explicit methodology.

Response 1:

(Please find uploaded Supplementary Table 1, "Representative molecular studies of human health", which we have put in great effort and gives an extensive summary of the studies and databases described in the manuscript.)

We thank the reviewer for this thoughtful and important critique. We agree that transparency in literature identification is essential, particularly given the breadth and interdisciplinary nature of this topic. While this manuscript is intended as a narrative and conceptual review rather than a formal systematic or scoping review, we have prepared Supplementary Table 1, to explicitly clarify the literature identification strategy used. In the Supplementary Table, we state:

"All of the studies was identified through a structured and transparent search strategy to minimize selection bias. Relevant studies were retrieved from PubMed, Web of Science, and Google Scholar using combinations of keywords including “metabolomics,” “metabolic profiling,” “health,” “healthy baseline,” “physiological homeostasis,” and “population cohort.” Priority was given to large-scale human cohort studies, representative methodological peer-reviewed papers, major talks/presentations personally given by the authors, and recent reviews that address metabolic variation in apparently healthy populations. This approach was informed by the general principles of the PRISMA guidelines, while allowing flexibility appropriate for a conceptual and integrative review.."

 

Comments 2:

Second, throughout the manuscript, the review primarily consists of a sequential listing of results from previous studies and extensive cohort studies without sufficient critical synthesis. Individual findings are described, but they are rarely interpreted in a way that yields deeper conceptual insight. Key questions—such as what fundamentally limits the definition of a metabolic health baseline, which conclusions are robust across platforms and cohorts, and which findings remain uncertain or contradictory—are not adequately addressed. As a result, the review reads more as an annotated bibliography than as a critical evaluation of the field.

Response 2:

We thank the reviewer for this insightful and important critique. We agree that a critical review should move beyond a descriptive summary of prior studies and provide integrative interpretation and conceptual synthesis. While one has to fully respect the description and discussions written in each peer-reviewed publication, iin this revision, we have added further analyses and critical synthesis of each study. Please find these in Suppl Table 1, where we provide columns listing "Cohort samples", "Objectives", "Key publications" "Constraints", "Key Advantages" and "Methodology". From these categories, the reader will be able to find a study that fits one's criteria.

 

With regards to the "Key Questions", we have added some general text that highlights the need to consider these (Page 12; Lines 493-512):

"Finally, beyond technical and design-related challenges, defining metabolic health baselines is constrained by fundamental conceptual limitations. Traditional approaches often assume a static notion of “normal” metabolism, whereas metabolic homeostasis is inherently dynamic and continuously shaped by age, lifestyle, environmental exposures, and physiological adaptation. Consequently, a single static baseline may fail to capture meaningful variations in metabolic health across time and contexts.

Furthermore, metabolic profiles are highly context-dependent, reflecting the integrated effects of multiple interacting determinants rather than isolated factors. This context dependency complicates efforts to define universal reference ranges applicable across life stages and populations. Defining “healthy” metabolism is particularly challenging across the life course, as metabolic requirements and normative ranges differ substantially between childhood, adulthood, aging, and transitional states such as pregnancy or menopause.

Collectively, these limitations highlight that defining metabolic health baselines is constrained not only by data availability but also by how measurements are generated, standardized, and interpreted over time, reinforcing the need for dynamic, stratified, and context-aware reference frameworks."

 

Comments 3:

Third, several core methodological issues are repeatedly acknowledged but not meaningfully analyzed. The manuscript emphasizes the need for large-scale and longitudinal data, yet does not critically review how such data can be generated and integrated in practice. Essential topics—such as long-term quality control, reproducible quantification across years, harmonization of data from different analytical platforms, and the separation of biological variability from measurement noise—are mentioned only in passing. Without addressing these issues from both measurement and data-analytical perspectives, the call for “larger datasets” remains generic and non-actionable.

Response 3:

We thank the reviewer for this important methodological critique. We agree that calls for larger and longitudinal datasets must be accompanied by a critical discussion of how such data can be generated, harmonized, and interpreted in practice. In response, we have substantially strengthened the discussion of methodological and analytical challenges, particularly throughout Sections 3 and 4.

• For new paragrpahs included, please see Page 12, Lines 493-512.
• (Another three papragraphs that overlaps with our response to Comment 5) Page 10, Line 426 to Page 11, Line 466)

 

Comments 4:

Fourth, the discussion of analytical platforms is oversimplified. Mass spectrometry is effectively limited to LC–MS and GC–MS, while other established approaches, such as CE–MS, which are particularly relevant for highly polar and ionic metabolites central to core metabolism, are entirely omitted. This contributes to an artificial NMR-versus-MS dichotomy and further limits the methodological completeness of the review.

Response 4:

Thank you for kindly pointing this out. We have added sentences pertaining to CE-MS on Page 9, Lines 375-379. With regards to the Reviewer's comment on "an artificial NMR-versus-MS dichotomy", our manuscript clearly states that (Page 9, Lines 385-396):

"To standardize the "health baseline" detection platform, the only way is to combine different detection platforms. According to previous studies, the metabolites detected by NMR and MS only have minimal overlaps [95,96]. Combining the two can increase the coverage of metabolites, thereby constructing a more complete and better-standardized health baseline, not only because the combination of NMR and MS expands the types of metabolites being detected but also due to the algorithms to detect unknown metabolites [97]. Currently, classic methods such as SUMMIT MS/NMR and NMR–MS Translator algorithms can identify the metabolites of substances without the need for compound purification [98,99]. In the case of such algorithms and the combination of multiple detection platforms, not only can the accuracy of the metabolic spectrum be improved, but also the construction of the health baseline becomes more complete, reliable, and biologically significant." 

Therefore, the paragraph concludes with stating the importance of a combination of both methodologies, not closing with an aritificial NMR versus MS dichotomy.

 

Comments 5:

Fifth, although artificial intelligence and machine learning are repeatedly highlighted as key future enablers, no concrete examples are provided of AIーbased metabolomics advancing to routine clinical or real-world implementation. The challenges specific to metabolomic data—such as confounding by lifestyle, diet, and short-term physiological variation; overfitting in cross-sectional designs; and translation to low-cost, deployable assays—are not critically examined. Consequently, the AI discussion remains aspirational and indistinguishable from that found in many existing reviews.

Response 5:

We thank the reviewer for this important and constructive critique. We agree that discussions of artificial intelligence and machine learning in metabolomics should move beyond aspirational statements and have provided concrete examples (Page 10 Line 427 to Page 121, Line 467):

"In terms of current developments, algorithmic interpretation of metabolomic data is operationalized in clinical workflows with measurable real-world impact. The Collaborative Laboratory Integrated Reports (CLIR) platform uses multivariate pattern recognition to integrate tandem MS profiles with covariate adjustments, replacing traditional cutoff values with continuous disease-likelihood scores. CLIR operationalizes multivariate post-analytical interpretation at scale, producing continuous likelihood scores rather than single cutoffs (e.g., hundreds of millions of post-analytical scores reported). Hall et al. report measurable improvements in screening performance (e.g., improved PPV and reduced false positives) for disorders of propionate, methionine, and cobalamin metabolism when CLIR tools and second-tier testing are used [118]. Similarly, NMR-based lipoprotein profiling (NMR LipoProfile) has received FDA 510(k) clearance for cardiovascular risk assessment, with millions of tests performed [119].

In addition, there are research tools that address the annotation bottleneck. Because the vast majority of small molecules lack experimental reference spectra, machine learning (ML) spectrum-prediction and in-silico annotation methods are increasingly necessary to scale identification beyond library matches. Graph transformer architectures (MassFormer) predict MS/MS spectra from molecular structures, enabling in silico library expansion beyond existing reference databases [120]. Complementary approaches using fragmentation tree analysis with ML-predicted molecular fingerprints (SIRIUS/CSI:FingerID) achieve substantially higher identification rates than spectral matching alone on benchmark datasets, without requiring spectral library matches [111,121].

Metabolomic profiles are sensitive to diet, circadian timing, and other short-term physiological states. This real-time-, sensitive- and direct-readout advantage may also lead to confounding data due to lifestyle, diet and short-term physiological variation. For technical variation, QC-anchored ML normalization methods (SERRF, hRUV) demonstrate practical reductions in batch effects [122,123]. SERRF, benchmarked against 15 competing methods across three large studies, reduced technical coefficients of variation to 5% RSD, enabling cross-study comparability [122]. For biological confounding, recent work demonstrates that lifestyle factors can be measured objectively as metabolite biomarkers: ML-validated poly-metabolite scores differentiate diet composition (80% vs. 0% ultra-processed food) within-individuals in randomized crossover trials [124]. ML-estimated circadian phase from plasma metabolomics achieves 0.45–0.60 hour accuracy under controlled conditions, enabling temporal standardization in future studies [125]. When lifestyle factors are strongly aligned with outcome (limited positivity) or lie on the causal pathway, statistical adjustment is not identifiable and can remove true biological signal. This is a fundamental constraint in causal inference. TRIPOD+AI and PROBAST+AI provide reporting guardrails for defensible model adjustment and when adjustment risks bias [126,127]."

 

 

Comments 6:

Finally, the manuscript's conceptual contributionーis limited. Core ideas such as the dynamic nature of health, the role of metabolomics as a functional readout, and the importance of large cohorts and AI integration have been extensively discussed in prior literature. The manuscript does not clearly articulate how its perspective differs from or advances beyond existing reviews, nor does it offer a novel framework or set of principles that would justify publication as a standalone synthesis.

Response 6:

We thank Reviewer 2 for providing this higher-level critique on our manuscript. As rightly mentioned, indeed core ideas such as "the dynamic nature of health, the role of metabolomics as a functional readout, and the importance of large cohorts and AI integration" have been touched upon in other works. However, we find that no one has truly integrated these ideas - along with extensively gathering and highlighting the features of previous peer-review studies. Hence, with all due respect, we feel confident that this review manuscript addresses a gap in knowledge.

 

We are grateful for the comments:

(Reviewer 1) "The authors bridge metabolomic research with precision medicine and population health, addressing a critical gap in understanding dynamic physiological homeostasis beyond disease absence."

(Reviewer 3) "This [sic] authors have presented an extremely well thought out review of an important topic for Metabolomics, i.e. the establishment of a possible baseline for comparison and detection of abnormalities."

(Reviewer 4); and "This manuscript describes issues regarding an interesting and ambitious subject, which is the definition of a “health baseline” and the potential application of metabolomics to this aim. The progress obtained with the analysis carried out in samples from biobanks, large-scale longitudinal cohorts, as well as the integration of data from different platforms using bioinformatic tools is thoroughtly described and discussed. A comparison of pros and cons is presented in Table 1. Major challenges are also properly discussed."

 

Thank you very much for signicficantly helping us improve our manuscript.

Reviewer 3 Report

Comments and Suggestions for Authors

This authors have presented an extremely well thought out  review of an important topic for Metabolomics, i.e. the establishment of a possible baseline for comparison and detection of abnormalities.  The manuscript is very well referenced, utilizing both the original papers where appropriate, and recent developments that will be of interest to the reader.  There conclusions are insightful.

The manuscript is logically organized, well thought out, and well written.

There are some inconsistencies, typos, word choice, and formatting irregularities that need to be addressed.

Specifically:

Pg. 3 line 114 typo/period mid sentence?

This review aims to discuss the role of metabolomics in possibly defining a health baseline. by primarily addressing the following aspects: current advances in metabolomics within health research; existing challenges; ongoing large-scale projects and research initiatives worldwide; and prospects.

 

Line 121 "...while informing understanding..." is awkward.  Perhaps "...while expanding our understanding..." would be better?

 

Line 136 to 137 the font type/size/spacing changes mid paragraph?  Is this a PDF irregularity or a formatting issue?

This happens again on line 147 and switched back again on line 160.

Line 169 switches again to the smaller font, and back again to the larger font 182. 

Line 184 small, line 197 back to large. Line 252, etc. This is very distracting and should have been checked before submission to reviewers.  This happens all over the document.

 

Line 241 "...Rothschild et al [64]." not formatted consistently with rest of document.

Line 252 "It goes without saying..." begs "well then why say it".  Recommend remove colloquialism and simply start with "Daily habits such as..." 

Line 301-314 - recommend authors remove "race" and instead keep with words such as "groups" etc., and emphasize the likely cultural differences influencing finds, e.g. possibly widely different dietary differences as opposed to implications regarding genetic differences.  Unless there are conclusive control publications showing clear genetic based metabolomic observable differences between groups that this reviewer is not aware of.  Removing "race" avoids any associated implications/questions/etc.

Table 1 mentioned first on pg 8 but does not appear until middle of pg9?  Could this table be moved up for more convenience to the readers?

Line 341 contradicts information in lines 333-336.

"...the detection limit of NMR is μM or nM, and it can only detect approximately 50- 200 types of metabolites [81]."

Also:

This result depends on many factors such as the library used, the source materials, type of instrumentation used, etc.

Recommend changing detection limit to "practical detection limit" has NMR does not have a detection limit, it has a time limit for the user to decide when the costs exceed the detection etc.

Line 344 "...sample volume of 30-600 μl..." recommend adding something to the effect "depending on the NMR probe used" as the sample volume is not as variable as indicated.

Line 377-382 seems like a copy/paste error occurred and the sentence(s) has been damaged for formatting and content?  Regardless the sentence is overly long and can be improved.  Perhaps a period after "...NMR data." and start the next sentence "This, combined with ....may overcome..." etc.

"The power of artificial intelligence will certainly go far beyond this, as it is believed that there will be artificial intelligence that unifies MS and NMR data, which combined with multi-omics data will overcome the challenges of data heterogeneity and obtaining a more com- prehensive metabolic view, constructing a dynamic, personalized, and systematic health reference framework."

Table 1 pg.9 needs to be consistent with text and references...can NMR identify dozens or 10s of thousands of metabolites?  Quantitation:  are the authors aware of what it takes to get accurate/precise quantitation via NMR?  Might want to mention as new users can be extremely demanding otherwise.

Line 387 damaged sentence or formatting?

"...healthy individu- als Noting that the majority..."

Line 392 - typo? "An good example would be the eye..." recommend "A good..."

Line 453 why is Age capitalized?  "...associated with gender differences (for example, serotonin has a higher content in females): Age was associated..."

Line 471 

It is worth noting the Trinity Student Study (TSS) as one that approaches to be a truly "healthy cohort" as much as possible.

This is very strange sentence structure and not sure what they authors are trying to emphasize.  May consider re-writing to be a bit more concise/clearer?

Line 538 seems to be a new section, maybe a section title would be helpful to readers?  Something like “Large Scale Projects” or “National BioBanks”?

Line 564 “…etc have not been listed. 

Etc needs to be consistently formatted throughout, recommend “etc.”

Also recommend “… etc. have not yet been added to their capabilities.” Or equivalent restructuring.  May want to include their plans (if any?), and what it takes to have key compounds added?

Line 603 format of Conclusion title has a period?  Typo/formatting error?

Really like the first paragraph of the conclusion.  Nicely written and clear.

Lines 630-639 may want to look deeper into inter-instrument/facility comparisons.  Specifically, it’s far harder to run consistent/accurate/precise NMR that is commonly thought.  It is not a “black box” and actually requires a great deal of NMR knowledge to properly compare data sets.

 

Author Response

We are most deeply grateful to Reviewer 3's meticulous review of our manuscript, and first and foremost, deeply apologise for not proofreading the manuscript more thoroughly. We are happy to report that all points mentioned have been addressed. These have been given within the MS Word-formatted revised manuscript (Track Changes enabled).

The following are responses to Reviewer 3's comments (excluding correcting spellings, alternative phrases, labelling, etc - All of these have been addressed):

 

--------

(1) Pg. 3 line 114 typo/period mid sentence? - This has been corrected. Thank you.

(2) Line 121 "...while informing understanding..." is awkward.  Perhaps "...while expanding our understanding..." would be better? - This has been corrected. Thank you.

(3) Line 136 to 137 the font type/size/spacing changes mid paragraph?  Is this a PDF irregularity or a formatting issue?  / This happens again on line 147 and switched back again on line 160.  /  Line 169 switches again to the smaller font, and back again to the larger font 182. /  Line 184 small, line 197 back to large. Line 252, etc. This is very distracting and should have been checked before submission to reviewers.  This happens all over the document.

We ourselves were quite surprised to see the different fonts-sizes. In our case, we did not use the MDPI manuscript template, but submitted our own MS Word document (which is allowed). The resulting different fonts / different size of text was a result of the automatic formatting post-uploading of our original file. We are jhappy to tell you that we have adjusted all of the incorrect fonts-sizes, and this issue has been resolved. 

(4)  Line 241 "...Rothschild et al [64]." not formatted consistently with rest of document.

This has been corrected.

(5) Line 252 "It goes without saying..." begs "well then why say it".  Recommend remove colloquialism and simply start with "Daily habits such as..."  - This has been corrected. Thank you kindly.

(6)   Line 301-314 - recommend authors remove "race" and instead keep with words such as "groups" etc., and emphasize the likely cultural differences influencing finds, e.g. possibly widely different dietary differences as opposed to implications regarding genetic differences.  Unless there are conclusive control publications showing clear genetic based metabolomic observable differences between groups that this reviewer is not aware of.  Removing "race" avoids any associated implications/questions/etc. - 

Thank you for pointing out this most-sensitive matter, and happy to tell you that we have rectified this.

 

(7)  Table 1 mentioned first on pg 8 but does not appear until middle of pg9?  Could this table be moved up for more convenience to the readers?

Table 1 has been moved up to appear right after the call-out within the text.

 

(8)  Line 341 contradicts information in lines 333-336. /  "...the detection limit of NMR is μM or nM, and it can only detect approximately 50- 200 types of metabolites [81]."  /  Also: This result depends on many factors such as the library used, the source materials, type of instrumentation used, etc. /  Recommend changing detection limit to "practical detection limit" has NMR does not have a detection limit, it has a time limit for the user to decide when the costs exceed the detection etc.

In all cases, the acceptable concentrations for NMR has been corrected. 

 

(9)  Line 344 "...sample volume of 30-600 μl..." recommend adding something to the effect "depending on the NMR probe used" as the sample volume is not as variable as indicated.

We have corrected this as kindly advised.

 

(10)   Line 377-382 seems like a copy/paste error occurred and the sentence(s) has been damaged for formatting and content?  Regardless the sentence is overly long and can be improved.  Perhaps a period after "...NMR data." and start the next sentence "This, combined with ....may overcome..." etc.

"The power of artificial intelligence will certainly go far beyond this, as it is believed that there will be artificial intelligence that unifies MS and NMR data, which combined with multi-omics data will overcome the challenges of data heterogeneity and obtaining a more com- prehensive metabolic view, constructing a dynamic, personalized, and systematic health reference framework."

Thank you for spotting this. We have made the correction by first, deleting this entire sentence. then secondly, strengthening the section on AI by adding three new paragraphs -  (Page 10 Line 427 to Page 121, Line 467):

"In terms of current developments, algorithmic interpretation of metabolomic data is operationalized in clinical workflows with measurable real-world impact. The Collaborative Laboratory Integrated Reports (CLIR) platform uses multivariate pattern recognition to integrate tandem MS profiles with covariate adjustments, replacing traditional cutoff values with continuous disease-likelihood scores. CLIR operationalizes multivariate post-analytical interpretation at scale, producing continuous likelihood scores rather than single cutoffs (e.g., hundreds of millions of post-analytical scores reported). Hall et al. report measurable improvements in screening performance (e.g., improved PPV and reduced false positives) for disorders of propionate, methionine, and cobalamin metabolism when CLIR tools and second-tier testing are used [118]. Similarly, NMR-based lipoprotein profiling (NMR LipoProfile) has received FDA 510(k) clearance for cardiovascular risk assessment, with millions of tests performed [119].

In addition, there are research tools that address the annotation bottleneck. Because the vast majority of small molecules lack experimental reference spectra, machine learning (ML) spectrum-prediction and in-silico annotation methods are increasingly necessary to scale identification beyond library matches. Graph transformer architectures (MassFormer) predict MS/MS spectra from molecular structures, enabling in silico library expansion beyond existing reference databases [120]. Complementary approaches using fragmentation tree analysis with ML-predicted molecular fingerprints (SIRIUS/CSI:FingerID) achieve substantially higher identification rates than spectral matching alone on benchmark datasets, without requiring spectral library matches [111,121].

Metabolomic profiles are sensitive to diet, circadian timing, and other short-term physiological states. This real-time-, sensitive- and direct-readout advantage may also lead to confounding data due to lifestyle, diet and short-term physiological variation. For technical variation, QC-anchored ML normalization methods (SERRF, hRUV) demonstrate practical reductions in batch effects [122,123]. SERRF, benchmarked against 15 competing methods across three large studies, reduced technical coefficients of variation to 5% RSD, enabling cross-study comparability [122]. For biological confounding, recent work demonstrates that lifestyle factors can be measured objectively as metabolite biomarkers: ML-validated poly-metabolite scores differentiate diet composition (80% vs. 0% ultra-processed food) within-individuals in randomized crossover trials [124]. ML-estimated circadian phase from plasma metabolomics achieves 0.45–0.60 hour accuracy under controlled conditions, enabling temporal standardization in future studies [125]. When lifestyle factors are strongly aligned with outcome (limited positivity) or lie on the causal pathway, statistical adjustment is not identifiable and can remove true biological signal. This is a fundamental constraint in causal inference. TRIPOD+AI and PROBAST+AI provide reporting guardrails for defensible model adjustment and when adjustment risks bias [126,127]."

 

(12)   Table 1 pg.9 needs to be consistent with text and references...can NMR identify dozens or 10s of thousands of metabolites?  Quantitation:  are the authors aware of what it takes to get accurate/precise quantitation via NMR?  Might want to mention as new users can be extremely demanding otherwise.

Thank you for pointing this out. We have attached qualifiers to clarify the number of metabolites / sensitivity of the technique.

 

(13)  Line 387 damaged sentence or formatting? "...healthy individu- als Noting that the majority..." /   Line 392 - typo? "An good example would be the eye..." recommend "A good..."  /   Line 453 why is Age capitalized?  "...associated with gender differences (for example, serotonin has a higher content in females): Age was associated..."

We are most grateful for the detailed reviewing of our manuscript. All three have been corrected. 

 

(14)   It is worth noting the Trinity Student Study (TSS) as one that approaches to be a truly "healthy cohort" as much as possible. This is very strange sentence structure and not sure what they authors are trying to emphasize.  May consider re-writing to be a bit more concise/clearer?

We have re-written this sentence (Page 14, Lines 588-589):  "It  is worth noting the Trinity Student Study (TSS) as one that may be considered to approach the analysis of a truly "healthy cohort"."

 

(15)  Line 538 seems to be a new section, maybe a section title would be helpful to readers?  Something like “Large Scale Projects” or “National BioBanks”?

The mention of "large scale projects" or "national biobanks" begins from Section 4.  We have therefore re-worded this sentence to make it flow better with the entire text.

 

(16)   Line 564 “…etc have not been listed. / Etc needs to be consistently formatted throughout, recommend “etc.” / Also recommend “… etc. have not yet been added to their capabilities.” Or equivalent restructuring.  May want to include their plans (if any?), and what it takes to have key compounds added?

Thank you for kindly pointing this out. We have corrected "etc." in throughout the text. With regards to this sentence, we have changed it as follows (Page 16, Lines 683-685):

"However, metabolites that are gaining increasing importance such as succinate, itaconate, etc. have not been identified amongst its list, and whether new compounds can be added post-analysis is not clear."

 

 

(17)  Line 603 format of Conclusion title has a period?  Typo/formatting error?

Thank you - A correction has been made.

 

(18)  Really like the first paragraph of the conclusion.  Nicely written and clear.

Our deepest thanks to Reviewer 3 for this most encouraging comment.

 

(19)   Lines 630-639 may want to look deeper into inter-instrument/facility comparisons.  Specifically, it’s far harder to run consistent/accurate/precise NMR that is commonly thought.  It is not a “black box” and actually requires a great deal of NMR knowledge to properly compare data sets.

Thank you for this comment. Being directly involved in NMR metabolomics, we are aware of the non-trivial specific practices that need to be performed in reproducible fashion to ensure compatibility with other platforms elsewhere. In parallel, we are currently witnessing development through the multivariable analysis software, which is able to take these files and have them compared in a less-subjective way. 

--------

 

(Additional Revisions Made)

Knowing that a text-based description of the various studies might not be the best way to introduce them, please find uploaded Supplementary Table 1, "Representative molecular studies of human health", which we have put in great effort and gives an extensive summary of the studies and databases described in the manuscript. The Supplementary Table uploaded provides detailed information,. where we provide columns listing "Cohort samples", "Objectives", "Key publications" "Constriants", "Key Advantages" and "Methodology". 

 

 

 

Again, our sincerest thanks for kindly carefully reviewing our manuscript, and allowing us to significantly improve our work.

Reviewer 4 Report

Comments and Suggestions for Authors

This manuscript describes issues regarding an interesting and ambitious subject, which is the definition of a “health baseline” and the potential application of metabolomics to this aim.  The progress obtained with the analysis carried out in samples from biobanks, large-scale longitudinal cohorts, as well as the integration of data from different platforms using bioinformatic tools is thoroughtly described and discussed. A comparison of pros and cons is presented in Table 1. Major challenges are also properly discussed. Nonetheless, some concerns need to be addressed:

• It is not clear whether the manuscript reports the objectives of the Project Baseline Health Study (PBHS) or the general state-of-the-art.
• Even though some lines are devoted to the existence of diversity of physiological status that can be accepted as “healthy”, there is not an appropriate discussion on how the “health baseline” will be established, i.e. for age ranges, for different ethnic groups, for each sex or gender, etc.
• A table depicting the cohorts and main results, standing metabolites, contributing to the aim of a health reference framework would be very valuable.
• Definition of the main factors that would affect the reference healthy status and the acceptable ranges for given metabolites would be of interest.
• The requirement of consistent results between studies using samples from different cohorts is stated by the authors, however “consistent” may not be equivalent to “equal” when different ethnic groups are compared, as the authors themselves point out. This should be remarked.
• Certainly to determine the biomarker “healthy” contents is of primary concern, but the lack of standards for thousands of compounds as well as the prohibitive cost are however a significant constraint that requires the collaboration of a number of research groups. This should be pointed out in the manuscript.

Author Response

Comments 1:  It is not clear whether the manuscript reports the objectives of the Project Baseline Health Study (PBHS) or the general state-of-the-art.

Response 1:

We thank the reviewer for this important clarification request. We have revised the manuscript to state that this review addresses the general state-of-the-art in defining metabolic health baselines, rather than reporting the objectives of a single initiative such as the Project Baseline Health Study. The objectives of the PBHS study is given to show that, (although the namesake contains the word "baseline") the actual study does not specifically aim to find the definition of health. We did leave the PBHS paragraph there, as it serves as a good example of the various efforts given in considering the importance of the metrics for health. The paragraph now reads (Page 3, Line 100 - Page 4, Line 118):

"To this end, the Project Baseline Health Study (PBHS) serves as a notable example [5]. The objectives of the study  are “(i) to develop the requisite tools and technologies, (ii) evaluate the use of sensor technologies, (iii) create a dataset encompassing a wide spectrum of phyotypic measures, (iv) measure the phenotypic diversity of the participants, and (v) share this vast data and create an example of open science” [40]. Currently, this research has.....  ...... Although certainly encompassing data from all individuals (healthy and unhealthy),  studies such as the systematic integration of metabolic profiles in a healthy state are not specified as its objectives, and therefore the project does not directly answer the question that we have stated above."

--------- 

With your permission, we address Comments 3 first (then Comments 2, 4, and 5 together):  A table depicting the cohorts and main results, standing metabolites, contributing to the aim of a health reference framework would be very valuable.

Response 3:

(Please find uploaded Supplementary Table 1, "Representative molecular studies of human health", which we have put in great effort and gives an extensive summary of the studies and databases described in the manuscript.)

We thank Reviewer 3 for this very helful suggestion. The Supplementary Table uploaded provides detailed information,. where we provide columns listing "Cohort samples", "Objectives", "Key publications" "Constriants", "Key Advantages" and "Methodology". 

 

-------

Comments 2:     Even though some lines are devoted to the existence of diversity of physiological status that can be accepted as “healthy”, there is not an appropriate discussion on how the “health baseline” will be established, i.e. for age ranges, for different ethnic groups, for each sex or gender, etc.

Comments 4:  Definition of the main factors that would affect the reference healthy status and the acceptable ranges for given metabolites would be of interest.

Comments 5: The requirement of consistent results between studies using samples from different cohorts is stated by the authors, however “consistent” may not be equivalent to “equal” when different ethnic groups are compared, as the authors themselves point out. This should be remarked.

Response 2, 4, and 5:

We agree with the reviewer that defining health baselines requires explicit stratification across demographic and biological dimensions. It would be ideal to be able to provide the acceptable ranges of metabolite concentrations for healthy individuals. And indeed the Reviewer is correct in pointing out that "consistency" and "equivalency" of measurements are not the same, especially when one takes into account the wide variations depending on the individual's age, gender, ethnicity, etc. In fact, it is exactly these variations (and variabilities) that were discussed in our manuscript that warrants a well-designed study (or a collaboration of studies after harmonisation of procedures, data, processing, etc).

 In response, we have expanded the discussion to emphasize these points. Please see:

(i) Page 12, Lines 494-512:

"Finally, beyond technical and design-related challenges, defining metabolic health baselines is constrained by fundamental conceptual limitations. Traditional approaches often assume a static notion of “normal” metabolism, whereas metabolic homeostasis is inherently dynamic and continuously shaped by age, lifestyle, environmental exposures, and physiological adaptation. Consequently, a single static baseline may fail to capture meaningful variations in metabolic health across time and contexts.

 Furthermore, metabolic profiles are highly context-dependent, reflecting the integrated effects of multiple interacting determinants rather than isolated factors. This context dependency complicates efforts to define universal reference ranges applicable across life stages and populations. Defining “healthy” metabolism is particularly challenging across the life course, as metabolic requirements and normative ranges differ substantially between childhood, adulthood, aging, and transitional states such as pregnancy or menopause.

Collectively, these limitations highlight that defining metabolic health baselines is constrained not only by data availability but also by how measurements are generated, standardized, and interpreted over time, reinforcing the need for dynamic, stratified, and context-aware reference frameworks."

(ii) And in the Conclusion, Page 16, Lines 752-765:

"At the same time, technical and analytical heterogeneity remains another obstacle to standardization. Differences between analytical platforms, particularly NMR- and MS-based metabolomics, limit cross-study comparability and complicate the establishment of universal reference frameworks. Rather than viewing these platforms as competing alternatives, the evidence strongly supports their integration. Multi-platform strategies, combined with advanced computational tools for metabolite identification and data harmonization, offer a path toward broader metabolome coverage and improved robustness. This is not simply a data management issue, as there are respective instrumentation-based, quality-control-related harmonization matters to be sorted as well. Moreover, the consistency in measurements and/or the equivalency of measured data points should be contextual, with age, gender, ethnicity, geography, and culture-influenced diet taken into account. Equally important is the recognition that metabolic baselines are biofluid- and compartment-specific, necessitating standardized protocols and metadata to ensure biological interpretability across tissues and sampling contexts."

-------

Comments 6: Certainly to determine the biomarker “healthy” contents is of primary concern, but the lack of standards for thousands of compounds as well as the prohibitive cost are however a significant constraint that requires the collaboration of a number of research groups. This should be pointed out in the manuscript.

Response 6: 

We fully agree with the reviewer that defining “healthy” biomarker contents faces substantial practical constraints, and have added this in our Conclusion. With regards to metabolite standards, (i) (although not-perfect and not all human) the Human Metabolome Database (HMDB) contains 220,945 metabolite entries with MS and NMR spectra, and (ii) various multivariate analysis software is continuously updating its database by incorporating more and more compounds, and therefore we feel that there is sufficient coverage in terms of the metabolites. (As probably intended by the Reviewer), it is the standard levels (absolute concentrations or relative amounts) amongst each healthy group that needs to be determined, and this will be achievable only through processing the data. Page 18, Lines 779-787 now read:

"Without doubt, large-scale biobanks and population cohorts now provide unprecedented opportunities to address these challenges going forward. Although most were not originally designed to define health baselines, their scale, diversity, and increasing availability of multi-omics and longitudinal data make them valuable resources. The process is also naturally collaborative interdisciplinarily and internationally (and hence may require significant initial investment), but will bear fruit as a catalogue of multi-omic standards. When combined with artificial intelligence and systems-level data integration, these datasets can support the construction of stratified, context-aware metabolic reference models that accommodate both individual variability and population heterogeneity."

------

 

We thank again Reviewer 4 for kindly helping us to improve our manuscript.

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The topic addressed by this review—defining a metabolic baseline of health using metabolomics—is timely and potentially important. The manuscript demonstrates that the authors are aware of relevant cohort studies and analytical platforms. Despite the breadth of material covered, the manuscript lacks the depth, focus, and conceptual coherence expected of a review by specialists in metabolomics-based cohort studies. As a result, the manuscript does not meet the standards of a rigorous academic review.

Major Concerns

1. Lack of Focus and Conceptual Coherence

Although the stated aim of the review is to discuss metabolomics in the context of cohort studies, the manuscript repeatedly diverges from this focus. Technical topics are introduced in a largely generic manner, without clear anchoring to the specific requirements, constraints, and design challenges of cohort-based metabolomics. Consequently, the narrative becomes fragmented, leaving readers without a clear understanding of the review's central message.

2. Misalignment Between Technical Discussion and Cohort Study Needs

A major weakness of the manuscript is the misalignment between the technical topics discussed and the methodological challenges that are actually critical in cohort studies.

• For example, the non-destructive nature of NMR is emphasized as a major advantage. However, the cohorts discussed in this review predominantly rely on consumable biofluids (e.g., serum, plasma, urine), where non-destructiveness is not a decisive benefit.
• In contrast, key NMR advantages that are highly relevant to long-term cohort studies—such as quantitative stability, low batch effects, and long-term reproducibility—are not sufficiently highlighted or critically discussed.
• Similarly, MS-based metabolomics is described mainly in terms of sensitivity and metabolite coverage, while well-known limitations in cohort settings (batch effects, instrument drift, long-term reproducibility) are not systematically analyzed. The discussion does not adequately focus on how these limitations are addressed through quality control strategies, normalization, and data harmonization.

As a result, the technical discussion remains generic and insufficiently tailored to cohort-based research.

3. Inappropriate Parallel Treatment of Genomic and Metabolomic Data

The manuscript often treats genomic and metabolomic data as conceptually equivalent and directly comparable. This is a fundamental conceptual problem.

• Germline genomic variants are largely stable over time and can be reliably captured with a single measurement.
• In contrast, metabolite levels are highly dynamic and sensitive to diet, season, circadian rhythms, and other environmental factors.

A metabolomics-centered cohort review should therefore focus on metabolite-specific challenges, such as controlling confounders, standardizing sampling conditions, handling repeated measurements, and interpreting temporal variability. Instead, the manuscript includes detailed genome-driven explanations (e.g., specific nucleotide substitutions) that are disproportionate to the metabolomics focus and distract from the central challenges unique to metabolomics.

4. Imbalance in Level of Detail

The level of detail across topics is inconsistent. In some sections, highly specific genomic details are provided, while metabolite-level changes, pathway-level interpretations, and cohort-specific metabolomics design issues are discussed only superficially. This imbalance further weakens the review's coherence and suggests that the manuscript lacks a clear hierarchy of importance among topics.

5. Structural Issues and Deterioration After Revision

The newly added or revised sections intended to address reviewer comments do not resolve the core issues outlined above. Instead, they further contribute to topic diffusion, making the manuscript's overall focus even less clear. Rather than sharpening the narrative, these additions reduce the review's clarity and academic quality.

While the general topic is appropriate and the manuscript demonstrates awareness of relevant literature, the current version lacks the conceptual depth, methodological focus, and integrative analysis required for a high-quality review article. To be suitable for publication, the manuscript would need to be fundamentally restructured around metabolomics-specific challenges in cohort studies, explicitly contrasting these with genomic cohort studies and critically analyzing current solutions and remaining limitations.

Given the extent of these issues, I recommend rejecting them rather than further revising them.

Comments on the Quality of English Language

In addition to the conceptual and structural concerns outlined above, the overall quality of the English language requires substantial improvement. While the manuscript is generally understandable, it contains frequent ambiguous references (e.g., overuse of “it” and “this” without clear antecedents), long and loosely structured sentences, and imprecise phrasing. These issues often obscure the intended meaning and make complex arguments difficult to follow.

Importantly, the language problems are not limited to minor grammatical errors that could be resolved by routine copyediting. Rather, the issues appear to reflect deeper problems in logical flow and sentence construction, suggesting that a higher level of professional scientific editing would be necessary to bring the manuscript to a publishable standard.

Given that this is a review article, where clarity, precision, and conceptual organization are particularly critical, the current level of English further exacerbates the manuscript's lack of focus and coherence. Addressing these language issues would require extensive rewriting rather than superficial correction.

Author Response

Dear Academic Editor:

Thank you very much for accepting our manuscript for publication. We have addressed the changes that you (AcE) have recommended. Please find the details in our Cover Letter.

- It therefore follows that we will not be addressing the comments and suggestions by Reviewer 2.

Our deepest thanks again for the opportunity to improve our manuscript, and wishing you only the very best in all your future endeavours.

 

Sincerely,

 

K.H. Mok