Metabolic Landscape of Endometrial Cancer: Insights into Pathway Dysregulation and Metabolic Features

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This manuscript explores tissue-based metabolic reprogramming in endometrial cancer using untargeted LC–HRMS metabolomics combined with pathway analysis and machine-learning-based feature selection. The topic is timely and relevant, and the focus on paired tumor and adjacent tissues is a clear strength compared with serum-based studies. The manuscript is generally well written, logically structured, and the data are presented clearly. However, several issues should be addressed before the work can be considered for publication:

The study includes 10 paired samples, which is understandable given the use of tissue material but the authors should comment how this limits statistical power and how does in affect the strength of claims regarding diagnostic performance.

ROC curves and AUC values were evaluated using the same dataset that was also used to select these six metabolites selected by Random Forest and SVM. When feature selection and performance testing are done on the same data, the resulting AUC values tend to be overly optimistic. Without cross-validation or an independent validation set, the true diagnostic performance cannot be reliably assessed. Authors should comment on that.

The study presents coherent and biologically plausible findings. With more cautious interpretation and clearer methodological detail and correct claims regarding biomarker performance, the manuscript could become suitable for publication.

Author Response

General Response

We sincerely thank the Editor and the reviewers for their thorough evaluation and constructive comments on our manuscript. All the insightful suggestions that have helped us to improve the manuscript.

In response to the reviewers’ comments, we have carefully revised the manuscript to provide a more cautious and accurate interpretation of our findings. In particular, we have explicitly acknowledged the limitations related to sample size and the exploratory nature of the machine-learning-based analyses, clarified the scope of claims regarding diagnostic performance, and expanded the discussion of methodological limitations and future validation strategies. We have also revised the Introduction and Conclusion sections to better contextualize our study within the broader challenges of metabolomics-based biomarker discovery.

All comments raised by the reviewers have been addressed in detail below in red. Corresponding revisions have been made in the manuscript, and we believe that these changes have significantly strengthened the clarity, rigor, and overall quality of the work.

 

Point-by-point response

 

Reviewer#1

 

Coments 1: This manuscript explores tissue-based metabolic reprogramming in endometrial cancer using untargeted LC-HRMS metabolomics combined with pathway analysis and machine-learning-based feature selection. The topic is timely and relevant, and the focus on paired tumor and adjacent tissues is a clear strength compared with serum-based studies. The manuscript is generally well written, logically structured, and the data are presented clearly.

 

Response 1: We sincerely thank the reviewer for the positive and encouraging comments on our manuscript. These constructive comments have been very helpful in improving the quality and clarity of the manuscript and have guided us in further refining the presentation of our work.

 

Coments 2: However, several issues should be addressed before the work can be considered for publication: The study includes 10 paired samples, which is understandable given the use of tissue material but the authors should comment how this limits statistical power and how does in affect the strength of claims regarding diagnostic performance.

 

Response 2: We thank the reviewer for this important comment. We acknowledge that the inclusion of 10 paired tumor and adjacent tissue samples represents a relatively small cohort, which inevitably limits statistical power and precludes definitive conclusions regarding diagnostic performance, such as sensitivity, specificity, and generalizability.

Importantly, the primary aim of this study is not to establish or validate a diagnostic model, but to explore tissue-level metabolic reprogramming in endometrial cancer and to identify candidate metabolites and dysregulated pathways through an exploratory metabolomics framework. The use of paired samples was intentionally chosen to minimize inter-individual variability and enhance the robustness of within-patient comparisons, thereby strengthening biological inference despite the modest sample size.

In response to this comment, we have revised the manuscript to explicitly acknowledge the limitations related to sample size and to temper claims regarding diagnostic performance. We now emphasize that the reported classification performance should be interpreted as exploratory and hypothesis-generating, and that future validation in larger, independent cohorts will be required.

 

Coments 3: ROC curves and AUC values were evaluated using the same dataset that was also used to select these six metabolites selected by Random Forest and SVM. When feature selection and performance testing are done on the same data, the resulting AUC values tend to be overly optimistic. Without cross-validation or an independent validation set, the true diagnostic performance cannot be reliably assessed. Authors should comment on that.

 

Response 3: We thank the reviewer for raising this important methodological concern. We agree that when feature selection and performance evaluation are performed on the same dataset, the resulting ROC curves and AUC values may be overly optimistic, and that, in the absence of cross-validation or an independent validation cohort, the true diagnostic performance cannot be reliably assessed. We fully acknowledge this limitation.

In the present study, machine-learning approaches were applied primarily for exploratory feature discovery and prioritization, rather than for establishing or validating a diagnostic model. Accordingly, we have revised the manuscript to clarify that the reported ROC and AUC analyses are intended to illustrate the discriminative potential of the selected metabolites within this exploratory framework, rather than to claim clinically validated diagnostic performance.

We have now explicitly discussed this limitation in the revised Discussion section and emphasized that future studies incorporating cross-validation strategies and independent cohorts will be required to confirm the diagnostic utility of these candidate metabolites.

 

Coments 4: The study presents coherent and biologically plausible findings. With more cautious interpretation and clearer methodological detail and correct claims regarding biomarker performance, the manuscript could become suitable for publication.

 

Response 4: We appreciate the reviewer’s constructive summary and helpful guidance. In response, we have revised the manuscript to adopt a more cautious and precise interpretation of our findings, clarified the methodological scope and limitations of the machine-learning analyses, and corrected the wording related to biomarker performance to avoid overinterpretation. We believe that these revisions have substantially improved the clarity, rigor, and overall suitability of the manuscript for publication.

 

 

Reviewer 2 Report

Comments and Suggestions for Authors

Review Report

This review concerns the manuscript entitled “Metabolic Landscape of Endometrial Cancer: Insights into Pathway Dysregulation and Biomarker Discovery.” The study investigates metabolic alterations in endometrial cancer using untargeted metabolomics analysis of paired tumor and adjacent normal tissues, combined with pathway enrichment analysis and machine-learning–based biomarker discovery. Overall, the study is technically well executed, with a clear experimental design and appropriate analytical methods. The use of paired tissue samples represents a major strength, as it minimizes inter-individual variability and enhances data robustness.

Sample selection appears to be well controlled, reducing potential confounding effects. The metabolomics workflow and statistical analyses are clearly described. Consequently, this manuscript is relevant to the field and represents a valuable contribution to cancer biomarker discovery using metabolomics approaches.

However, several points require clarification and revision before publication. First, the authors should clearly acknowledge the limitations of both the metabolomics field in general and the present study in particular. Specifically:

Sample size limitation: The cohort size (n = 10 paired samples) is relatively small for robust biomarker discovery, particularly when machine-learning methods are applied. Despite the reported high diagnostic accuracy, there is a substantial risk of model overfitting.

Biomarker claims: The conclusion that the six identified metabolites are “promising diagnostic biomarkers” appears premature without validation in independent patient cohorts or in clinically relevant biofluids (e.g., plasma or serum). The clinical relevance and translational feasibility of these findings should be discussed more explicitly.

Conclusion section: The conclusion should be expanded to include recommendations for future studies aimed at validating the identified metabolites and pathways. In addition, the authors should clarify whether these metabolic signatures and pathways are novel in the context of endometrial cancer and discuss their potential utility as biomarkers.

Machine-learning limitations: The inherent limitations and challenges of machine-learning approaches, including overfitting, interpretability, and reproducibility, should be explicitly addressed.

Furthermore, it is recommended that the introduction include a brief discussion of the general limitations of metabolomics-based biomarker discovery. The following references may be helpful in framing these challenges and preventing overinterpretation of the results:

- Advances in mass spectrometry for metabolomics: Strategies, challenges, and innovations in disease biomarker discovery

- Metabolomics analysis for biomarker discovery: advances and challenges

- NMR-based metabolomics in human disease diagnosis: applications, limitations, and recommendations

Overall, the manuscript is above average compared with similar studies in the field and, after minor revisions and additions, would be suitable for publication.

 

 

Author Response

General Response

We sincerely thank the Editor and the reviewers for their thorough evaluation and constructive comments on our manuscript. All the insightful suggestions that have helped us to improve the manuscript.

In response to the reviewers’ comments, we have carefully revised the manuscript to provide a more cautious and accurate interpretation of our findings. In particular, we have explicitly acknowledged the limitations related to sample size and the exploratory nature of the machine-learning-based analyses, clarified the scope of claims regarding diagnostic performance, and expanded the discussion of methodological limitations and future validation strategies. We have also revised the Introduction and Conclusion sections to better contextualize our study within the broader challenges of metabolomics-based biomarker discovery.

All comments raised by the reviewers have been addressed in detail below in red. Corresponding revisions have been made in the manuscript, and we believe that these changes have significantly strengthened the clarity, rigor, and overall quality of the work.

 

 

Point-by-point response

 

Reviewer#2

 

Coments 1: This review concerns the manuscript entitled “Metabolic Landscape of Endometrial Cancer: Insights into Pathway Dysregulation and Biomarker Discovery.” The study investigates metabolic alterations in endometrial cancer using untargeted metabolomics analysis of paired tumor and adjacent normal tissues, combined with pathway enrichment analysis and machine-learning–based biomarker discovery. Overall, the study is technically well executed, with a clear experimental design and appropriate analytical methods. The use of paired tissue samples represents a major strength, as it minimizes inter-individual variability and enhances data robustness.

Sample selection appears to be well controlled, reducing potential confounding effects. The metabolomics workflow and statistical analyses are clearly described. Consequently, this manuscript is relevant to the field and represents a valuable contribution to cancer biomarker discovery using metabolomics approaches.

However, several points require clarification and revision before publication. First, the authors should clearly acknowledge the limitations of both the metabolomics field in general and the present study in particular.

 

Response 1: We thank the reviewer for the thorough and positive evaluation of our manuscript, including the recognition of the technical quality, paired tissue design, and robustness of the metabolomics workflow. We appreciate the assessment that this study represents a relevant contribution to metabolomics-based research in endometrial cancer. In response to the reviewer’s comment, we have revised the manuscript to more clearly acknowledge both the general limitations of metabolomics-based biomarker discovery and the specific limitations of the present study. These points are now explicitly addressed in the Introduction, Discussion, and Conclusion sections.

 

Coments 2: Sample size limitation: The cohort size (n = 10 paired samples) is relatively small for robust biomarker discovery, particularly when machine-learning methods are applied. Despite the reported high diagnostic accuracy, there is a substantial risk of model overfitting.

 

Response 2: We thank the reviewer for highlighting this important limitation. We acknowledge that the relatively small cohort size limits the robustness of biomarker discovery and increases the risk of overfitting when applying machine-learning approaches. These issues are now explicitly addressed in the Limitations section of the revised manuscript. We further clarify that machine-learning analyses were conducted in an exploratory manner for feature discovery and prioritization rather than for establishing validated diagnostic models, and we emphasize the need for validation in larger, independent cohorts.

 

Coments 3: Biomarker claims: The conclusion that the six identified metabolites are “promising diagnostic biomarkers” appears premature without validation in independent patient cohorts or in clinically relevant biofluids (e.g., plasma or serum). The clinical relevance and translational feasibility of these findings should be discussed more explicitly.

 

Response 3: We agree with the reviewer that definitive claims regarding diagnostic biomarkers are premature in the absence of validation in independent cohorts or clinically relevant biofluids. Accordingly, we have revised the manuscript to remove or temper statements referring to the six metabolites as promising diagnostic biomarkers. These metabolites are now consistently described as candidate metabolic features identified within an exploratory, tissue-based framework. In addition, we have expanded the Discussion and Conclusion sections to more explicitly address the clinical relevance and translational feasibility of these findings, emphasizing that future validation in larger, independent patient cohorts and in accessible biofluids such as plasma or serum will be required to establish diagnostic utility.

 

Coments 4: Conclusion section: The conclusion should be expanded to include recommendations for future studies aimed at validating the identified metabolites and pathways. In addition, the authors should clarify whether these metabolic signatures and pathways are novel in the context of endometrial cancer and discuss their potential utility as biomarkers.

 

Response 4: We thank the reviewer for this constructive suggestion. In response, we have expanded the Conclusion section to more clearly outline future research directions, including validation of the identified metabolites and pathways in larger, independent cohorts and functional studies to assess their biological relevance. We have also clarified the novelty of our findings by noting that while certain pathways, such as sphingolipid and glutathione metabolism, have been previously implicated in endometrial cancer, our study provides an integrated, tissue-level metabolomic perspective that refines and extends existing knowledge. Furthermore, we discuss the potential utility of these metabolic signatures as candidate features for biomarker development, emphasizing that their clinical relevance will require rigorous validation in future studies.

 

Coments 5: Machine-learning limitations: The inherent limitations and challenges of machine-learning approaches, including overfitting, interpretability, and reproducibility, should be explicitly addressed.

 

Response 5: We agree with the reviewer that machine-learning approaches have inherent limitations, including risks of overfitting, challenges in biological interpretability, and issues related to reproducibility. These limitations are now explicitly addressed in the Discussion and Limitations sections of the revised manuscript. We clarify that machine-learning methods in this study were applied in an exploratory manner for feature prioritization rather than for establishing predictive or diagnostic models, and we emphasize the need for independent validation to ensure robustness and reproducibility.

 

Coments 6: Furthermore, it is recommended that the introduction include a brief discussion of the general limitations of metabolomics-based biomarker discovery. The following references may be helpful in framing these challenges and preventing overinterpretation of the results:

- Advances in mass spectrometry for metabolomics: Strategies, challenges, and innovations in disease biomarker discovery

- Metabolomics analysis for biomarker discovery: advances and challenges

- NMR-based metabolomics in human disease diagnosis: applications, limitations, and recommendations

 

Response 6: We thank the reviewer for this helpful recommendation. In response, we have revised the Introduction to include a brief discussion of the general limitations and challenges of metabolomics-based biomarker discovery, including analytical variability, high-dimensional data interpretation, and the need for rigorous validation to avoid overinterpretation. The suggested references have been incorporated to appropriately frame these challenges within the context of the current study.