A Cytokine-Related Gene Signature for Pan-Cancer Prognostic Stratification and Malignant Phenotype Characterization

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The research topic holds significant clinical value, focusing on the prognostic role of cytokine-related genes in pan-cancer, which fills the gap left by studies on individual cytokines or limited signaling pathways that struggle to capture the systemic impact of cytokine networks. It provides a novel perspective for pan-cancer prognostic stratification and precision oncology.
The study design was rigorous, employing LASSO Cox regression to construct a 16-gene prognostic signature. Multi-level validation, including training and testing sets, pan-cancer and cancer-specific analyses, and the integration of clinical variables to build a nomogram, ensured the robustness and reproducibility of the results.
The multidimensional mechanism was comprehensively investigated through survival analysis, pathway enrichment, malignant phenotype association, and immunohistochemical validation, clearly elucidating the correlation between high-risk scores and malignant phenotypes such as epithelial-mesenchymal transition and cell cycle activation, providing robust support for the biological plausibility of prognostic features.

But the areas that need to be modified and suggestions are as follows:

1 The data source is single, relying solely on retrospective data from the TCGA database, lacking prospective clinical cohort validation, making it difficult to completely eliminate selection bias. It is recommended to supplement independent prospective cohort data or multi center validation to enhance clinical translational value.

2. Mechanism research focuses on association analysis and does not clarify the causal roles and molecular regulatory pathways of key genes (such as PANX1 and FRMD8) in the 16 gene signature. It is recommended to increase the analysis and research on causal relationships.

3. The description of immune infiltration analysis is brief and does not elaborate on the interaction between cytokine related risk scores and immune cell subsets (such as T cells and macrophages) in the tumor microenvironment. It is recommended to supplement quantitative analysis and correlation verification of immune cell infiltration.

4. Not considering the analysis of regulatory relationships between genes, it is recommended to add gene regulatory network analysis.

5. The correlation between this prognostic feature and existing clinical treatment options (such as immunotherapy and chemotherapy) has not been explored, and its guiding significance in treatment decision-making cannot be clarified. It is recommended to supplement subgroup analysis to evaluate the differences in response of high-risk/low-risk patients to different treatment options.

Author Response

Comment 1:
The data source is single, relying solely on retrospective data from the TCGA database, lacking prospective clinical cohort validation, making it difficult to completely eliminate selection bias. It is recommended to supplement independent prospective cohort data or multi-center validation to enhance clinical translational value.

Response:
We sincerely thank the reviewer for this insightful and important comment. We fully acknowledge that the present study is primarily based on retrospective transcriptomic data from TCGA, which may introduce potential selection bias and limit direct clinical generalizability.

However, TCGA represents  a large-scale, multi-institutional consortium comprising samples collected from numerous medical centers across different geographic regions. This multi-center composition partially mitigates single-center bias and enhances population heterogeneity. To further strengthen methodological rigor and reduce overfitting, we performed independent internal validation by randomly dividing the dataset into training and test cohorts. The cytokine-related risk score demonstrated consistent prognostic performance, stable hazard ratios, and reproducible survival stratification across multiple clinical endpoints (OS, DSS, and PFI) in both cohorts, as well as across different cancer types. These findings support the robustness and generalizability of the proposed signature within the TCGA framework.

We agree with the reviewer that independent prospective or multi-center cohort validation would further strengthen the translational value of this work. Nevertheless, the primary objective of this study was to establish a pan-cancer molecular framework and provide biologically grounded prognostic insights rather than to develop an immediately deployable clinical assay. Prospective validation in well-annotated clinical cohorts will be an important next step toward clinical translation.

In response to the reviewer’s suggestion, we have now explicitly clarified this limitation and future direction in the revised Discussion section. We appreciate this constructive comment, which has helped us better define the scope and clinical implications of our study.

Comment 2:
Mechanism research focuses on association analysis and does not clarify the causal roles and molecular regulatory pathways of key genes (such as PANX1 and FRMD8) in the 16-gene signature. It is recommended to increase the analysis and research on causal relationships.

Response: We sincerely thank the reviewer for this valuable comment. We agree that the present study primarily establishes robust association between the cytokine-related gene signature and clinical outcomes, malignant phenotypes, and pathway-level alterations, rather than directly demonstrating causal molecular mechanisms for individual genes such as PANX1 and FRMD8.

The primary objective of this work was to construct and validate a pan-cancer prognostic framework based on coordinated cytokine-related transcriptional programs. To support biological relevance, we integrated multiple layers of analysis, including survival modeling, pathway enrichment, malignant phenotype correlations (EMT and cell cycle activity), and protein-level validation using immunohistochemistry. These analyses consistently suggest that high-risk tumors are characterized by enhanced proliferative and inflammatory signaling programs.

With respect to PANX1 and FRMD8 specifically, prior studies have reported their involvement in inflammatory signaling, membrane channel activity, immune modulation, and tumor-associated pathways, supporting their potential functional relevance within cytokine-regulated tumor microenvironments. However, we acknowledge that definitive causal relationships and detailed molecular regulatory mechanisms require dedicated experimental studies, such as gene knockdown/overexpression models, pathway perturbation assays, and in vivo validation.

In response to the reviewer’s suggestion, we have now expanded the Discussion section to explicitly clarify that the current study identifies clinically relevant associations and generates mechanistically informed hypotheses, while functional validation of individual genes represents an important direction for future research.

We appreciate the reviewer’s constructive recommendation, which has helped us better define the mechanistic scope and future investigative directions of this work.

Comment 3:
The description of immune infiltration analysis is brief and does not elaborate on the interaction between cytokine-related risk scores and immune cell subsets (such as T cells and macrophages) in the tumor microenvironment. It is recommended to supplement quantitative analysis and correlation verification of immune cell infiltration.

Response: We sincerely thank the reviewer for this valuable suggestion. We acknowledge that the description of immune infiltration analysis in the original manuscript was relatively concise and did not extensively elaborate on interactions between the cytokine-related risk score and specific immune cell subsets such as T cells and macrophages. However, the primary objective of this study was to establish and validate a robust pan-cancer cytokine-related prognostic framework and to characterize its global biological associations. The immune-related findings presented in the manuscript were intended to provide supportive contextual interpretation rather than a comprehensive immune deconvolution analysis.

To address the reviewer’s concern, we have revised the Results and Discussion sections to clarify the scope of the immune analysis and to moderate our interpretation accordingly. We now explicitly state that the observed associations reflect correlative relationships and should be interpreted as hypothesis-generating rather than mechanistic evidence. We have also emphasized that detailed immune subset–specific analyses and functional validation represent important directions for future investigation.

We appreciate the reviewer’s insightful suggestion, which has helped us refine the discussion and more clearly define the boundaries and future perspectives of our work.

 

Comment 4:
Not considering the analysis of regulatory relationships between genes, it is recommended to add gene regulatory network analysis.

Response: We sincerely thank the reviewer for this constructive suggestion. We agree that gene regulatory network analysis could provide additional systems-level insight into potential regulatory relationships among the signature genes.

However, the primary objective of the present study was to construct and validate a clinically relevant cytokine-related prognostic signature at the pan-cancer level rather than to comprehensively delineate upstream regulatory circuitry or gene interaction networks. Incorporating full gene regulatory network modeling would substantially expand the analytical scope and move beyond the central aim of this work.

Instead of adding new network analyses, we have carefully revised the manuscript to moderate our mechanistic interpretation. Specifically, we have toned down statements that could imply direct regulatory relationships among the 16 genes and clarified that the observed associations are based on coordinated expression patterns rather than experimentally validated regulatory interactions. The Discussion section has been revised accordingly to emphasize that detailed gene regulatory network analysis represents an important direction for future investigations.

We appreciate the reviewer’s thoughtful recommendation, which has helped us refine the presentation of our findings and more clearly define the conceptual boundaries of the study.

Comment 5：

The correlation between this prognostic feature and existing clinical treatment options (such as immunotherapy and chemotherapy) has not been explored, and its guiding significance in treatment decision-making cannot be clarified. It is recommended to supplement subgroup analysis to evaluate the differences in response of high-risk/low-risk patients to different treatment options.

Response: We sincerely thank the reviewer for this important and clinically relevant suggestion. We agree that evaluating the relationship between the cytokine-related risk score and treatment response including immunotherapy and chemotherapy outcomes, would provide valuable translational insight. However, the TCGA pan-cancer dataset used in this study does not contain sufficiently detailed and standardized treatment response information across tumor types to allow reliable subgroup analysis of therapeutic outcomes. Treatment regimens, response criteria, and follow-up strategies vary substantially across cancer types and institutions, which limits the feasibility of conducting consistent and interpretable treatment-response stratification analyses within this framework.

The primary aim of this study was to establish and validate a pan-cancer prognostic signature and to characterize its biological associations rather than to evaluate predictive value for specific therapies. To avoid overstating clinical applicability, we have revised the Discussion section to clarify that the current findings support prognostic stratification but do not establish predictive value for treatment response. We now explicitly state that future studies using well-annotated therapeutic cohorts, particularly immunotherapy-treated populations, will be necessary to determine whether the cytokine-related risk score has predictive utility in guiding treatment decision-making.

We appreciate the reviewer’s constructive comment, which has helped us better define the translational scope and future directions of this work.

Reviewer 2 Report

Comments and Suggestions for Authors

Review Report

Title: “A cytokine-related gene signature for pan-cancer prognostic stratification and malignant phenotype characterization”

Authors: Shih-Chieh Chen, Kai-Fu Chang, Chien-Cheng Chao, Chung-Hsien Lin, Chih-Hsuan Chang, Ching-Chung Ko, Hui-Ru Lin, Chi-Jen Wu, Chien-Han Yuan, Sachin Kumar, Dahlak Daniel Solomon, Do Thi Minh Xuan, Neethu Palekkode, Ayman Fathima, Junanda Waikhom, Chih-Yang Wang, Yung-Kuo Lee, Hui-Pu Liu.

Comments:

In their paper “A cytokine-related gene signature for pan-cancer prognostic stratification and malignant phenotype characterization” Chen and colleagues aimed at investigating systematically cytokine-related genes across pan-cancer cohorts to develop and validate a robust prognostic risk score.

The article is a well-designed study, integrating tran-scriptomic data with survival outcomes. They constructed a cytokine-related gene signature that stratifies patients into distinct risk groups with significantly different clinical prognoses. Furthermore, they evaluated the distribution and prognostic performance of this signature across cancer types, assessed its clinical utility through nomogram modeling, and explored its biological relevance by correlating the risk score with malignant phenotypes and functional enrichment patterns. They analyzed Pan-cancer transcriptomic and clinical data and construct a cytokine-related prognostic signature using LASSO Cox regression. They stratified patients into high-risk and low-risk groups based on the derived risk score then evaluated prognostic performance in training and test cohorts and assessed biological relevance through survival analyses and pathway-level investigations. 16-gene cytokine-related signature was established that consistently stratified patients into distinct prognostic groups across multiple cancer types. High cytokine-related risk scores were significantly associated with unfavorable survival outcomes and were linked to enhanced cell cycle activity, epithelial-mesenchymal transition, and extracellular matrix remodeling. Integration of the risk score with clinical variables improved individualized survival prediction. Immunohistochemical analyses further confirmed increased protein expression of representative risk-associated genes, including PANX1 and FRMD8, in multiple tumor tissues compared with corresponding normal tissues

Taken together, the study is well conducted and adds substantial new knowledge regarding the cytokine-related prognostic signature. The findings of this article provide a comprehensive framework for understanding the prognostic and biological significance of cytokine dysregulation in cancer. Consequently, this could be beneficial for pan-cancer risk stratification and personalized clinical management.

Comments:

1. In Line 358, What is the source of the immunohistochemical staining data used in the article?

The authors have to add the reference to this data in this section.

 

2. In Section: 2.7 Immunohistochemical validation of PANX1 and FRMD8 expression in normal and malignant 355 tissues. It is not enough to show the difference of protein expression in normal tissue and corresponding cancer. Because the cancer arising from any tissue has different grades. It is better to show the correlation between PANX1 and FRMD8 protein expressions and tumor grade to get better idea about their prognostic role in cancer.

Therefore, this section isn’t enough for Immunohistochemical validation of PANX1 and FRMD8.

 

3. In Figure 7: There is no magnification scales on the immunohistochemical images. It is better to add it under each tissue core

 

4. There is a limitation of the used data to TCGA data mostly. It would be better if the authors could test other publicly available data like the Gene Expression Omnibus (GEO). Furthermore, the validation of the results would be better to be done using cohort of patients.

 

Author Response

Comment 1:
In Line 358, what is the source of the immunohistochemical staining data used in the article? The authors have to add the reference to this data in this section.

Response:
We thank the reviewer for this important comment. The immunohistochemical staining images used in this study were obtained from the Human Protein Atlas database. We have now clearly specified the data source in the revised Materials and Methods section and added the appropriate reference citation.

In the revised manuscript, we have now clearly specified the Human Protein Atlas as the data source in the corresponding section of the Materials and Methods and have added the appropriate reference citation. We appreciate the reviewer’s careful reading, which has helped improve the clarity and transparency of the manuscript.

Comment 2:
In Section 2.7, it is not sufficient to show differences in protein expression between normal and cancer tissues. Since cancers have different grades, it would be better to show the correlation between PANX1 and FRMD8 protein expression and tumor grade to better evaluate their prognostic role. Therefore, this section is not sufficient for immunohistochemical validation.

Response: We sincerely thank the reviewer for this valuable suggestion. We agree that evaluation of protein expression across tumor grades would provide additional insight into the potential prognostic relevance of PANX1 and FRMD8. However, the immunohistochemical data used in this study were obtained from the Human Protein Atlas, which provides representative tissue staining images but does not consistently include standardized tumor grade information or  sufficient sample-level quantitative data across cancer types to enable reliable grade-based statistical analysis. Therefore, a systematic correlation between protein expression levels and tumor grade could not be performed within the current framework.

We would like to clarify that the purpose of Section 2.7 was to provide supportive protein-level validation demonstrating differential expression between normal and tumor tissues, rather than to establish independent prognostic value at the protein level. To avoid overinterpretation, we have revised the text to tone down statements suggesting prognostic implications derived from IHC findings and now emphasize that these results serve as qualitative translational support for the transcriptomic signature.

We agree that future studies incorporating well-annotated clinical cohorts with matched tumor grade and outcome data would be valuable to further evaluate the association between PANX1 and FRMD8 protein expression and tumor aggressiveness.

We appreciate the reviewer’s constructive feedback, which has helped us clarify the scope and interpretation of our immunohistochemical analysis.

 

Comment 3:
In Figure 7, there are no magnification scales on the immunohistochemical images. It would be better to add them under each tissue core.

Response: We thank the reviewer for this careful observation. We agree that inclusion of magnification scales would improve figure clarity and reproducibility. The immunohistochemical images were obtained from the Human Protein Atlas, where representative images are provided without embedded scale bars in the downloadable figures. In the revised manuscript, we have added the original magnification information as provided by the database in the figure legend to ensure transparency. Where possible, we have also updated the figure to indicate magnification details consistently across panels.

We appreciate the reviewer’s suggestion, which has improved the quality and presentation of Figure 7.

Comment 4:
There is a limitation of the used data to TCGA data mostly. It would be better if the authors could test other publicly available data like the Gene Expression Omnibus (GEO). Furthermore, validation of the results using a patient cohort would be preferable.

Response: We thank the reviewer for this important comment. We acknowledge that the present study is primarily based on TCGA data. However, TCGA represents a large-scale, multi-institutional consortium that integrates samples collected from numerous centers and diverse patient populations, thereby reducing center-specific bias compared with single-institution datasets.

To enhance robustness, we performed independent internal validation by dividing the cohort into training and test sets and demonstrated consistent prognostic performance across both cohorts and multiple cancer types. In addition, the stability of hazard ratios and consistent survival stratification across endpoints further support the generalizability of the cytokine-related signature.

We agree that validation in external GEO datasets or prospective patient cohorts would provide additional clinical confirmation. However, given the pan-cancer framework and the heterogeneity of available GEO datasets in terms of platform, annotation completeness, and endpoint definition, systematic cross-cancer validation was not feasible within the scope of the current study.

We have now explicitly acknowledged this limitation in the revised Discussion and emphasized that future prospective and multi-center validations will be necessary to further establish clinical utility.

We appreciate the reviewer’s constructive suggestion, which has helped us clarify the scope and limitations of the study.

 

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

No comments.