Escribir un comentario

This review is the result of a virtual, collaborative live review discussion organized and hosted by PREreview on February 17, 2026, as part of the PREreview Champions Program 2026. The discussion was joined by 8 people: 2 facilitators from the PREreview Team, and 6 members of the PREreview 2026 Champions Program cohort. The authors of this review have dedicated additional asynchronous time over the course of two weeks to help compose this final report using the notes from the Live Review. We thank all participants who contributed to the discussion and made it possible for us to provide feedback on this preprint.

Summary

The study investigated whether a composite transcriptional signature that captures the interplay between glutamine metabolism and interferon-gamma (IFNγ) associated immune signalling could serve as an independent prognostic or predictive biomarker in melanoma. Specifically, the authors tested the hypothesis that tumors with heightened glutamine metabolic activity may exhibit weaker IFNγ-driven anti-tumor immune responses, reflecting a state of “metabolic-immune conflict”, and that this relationship might influence overall survival or response to anti-PD-1 therapy. To address this question, they developed two gene expression-based signatures grounded in established signalling pathways and prior literature, calculated enrichment scores using GSVA, and evaluated their associations with clinical outcomes. Prognostic performance was assessed in the The Cancer Genome Atlas skin cutaneous melanoma cohort (TCGA-SKCM; n = 469), while the predictive value for immunotherapy response was validated in an independent anti-PD-1 dataset from the Gene Expression Omnibus (GSE91061). Although the glutamine metabolism and IFNγ signatures were negatively correlated, supporting the biological premise of metabolic-immune antagonism, the composite signature did not demonstrate independent prognostic significance and failed to predict response to anti-PD-1 therapy, showing only a non-significant survival trend in the favourable group.

The topic is timely and biologically well-motivated, particularly given the increasing interest in metabolic immune interactions in the tumor microenvironment. These findings underscore the importance of rigorously testing biologically compelling hypotheses and highlight the value of transparently reporting negative results, thereby sparing other researchers from pursuing similar dead ends. While strengths of the study include the use of large independent cohorts and a robust statistical modelling framework, limitations include limited comparison with related studies and a deeper contextual discussion. Overall, the work contributes meaningfully to the field by demonstrating that a plausible metabolic-immune transcriptional signature does not translate into clinically actionable predictive value in melanoma. Moving forward, more granular methodologies, such as single-cell transcriptomics, may be warranted to capture the details this bulk-level analysis may have overlooked.

Feedback and list of major and minor concerns:

The use of publicly available cohorts and independent validation is commendable, and the willingness to report negative findings is scientifically valuable and contributes to transparency in biomarker research. However, the “metabolic conflict” hypothesis is clearly articulated, but the manuscript oversimplifies the metabolic landscape by focusing almost exclusively on glutamine metabolism. Melanoma metabolism is multifaceted (glycolysis, oxidative phosphorylation, lipid metabolism, etc.), and immune cell metabolic plasticity may weaken the explanatory power of a single-axis model. The authors may consider expanding justification for prioritizing glutamine over other metabolic pathways and discussing alternative metabolic axes or incorporate multi-pathway modeling. Also, they could clarify mechanistic links between transcriptional signatures and functional metabolic flux, include literature discussing metabolic heterogeneity and pathway cross-talk, and frame findings within this broader context. It is unclear whether gene selection bias influenced outcomes. Hence the authors could provide reproducible criteria/workflow for gene selection; Compare results using alternative curated or database-derived signatures. And consider pathway enrichment validation or bootstrapping.

1. Risk of Cell-Type Confounding in Bulk RNA-seq Data A central methodological concern is the reliance on bulk RNA sequencing, which averages gene expression across tumor, stromal, and immune compartments, as acknowledged in the Discussion. Differences in tumor purity, immune infiltration, and stromal content could confound the observed negative correlation between glutamine metabolism and IFNγ response. The apparent “metabolic-immune conflict” may thus reflect compositional differences rather than true biological antagonism. We suggest applying immune deconvolution algorithms (e.g., CIBERSORT, xCell, EPIC) to estimate cell-type proportions and adjust for them in multivariable models. Alternatively, restrict glutamine metabolism genes to those predominantly expressed in melanoma cells using published single-cell datasets. Sensitivity analyses stratified by tumor purity would further strengthen causal interpretation.

2. Incomplete Covariate Adjustment Although the study adjusted for age, sex, and tumor stage, other clinically relevant covariates were not considered. Variables such as comorbidities, body mass index (which may influence glutamine metabolism), performance status, and treatment-related factors could confound associations with survival or immunotherapy response. We suggest the author expand multivariable modeling to include these additional covariates where possible. We also recommend that the author explicitly discuss these other confounding variables and how they might impact the outcome/interpretation of the study.

3. Cohort Reduction and Potential Selection Bias Approximately one-third of The Cancer Genome Atlas melanoma cohort (469 to 313 patients) was excluded from survival analyses. The characteristics of excluded versus included patients were not described. If missing survival or clinical data are non-random (e.g., related to stage or sample site), this may introduce bias and compromise generalisability. Providing a comparison table of baseline characteristics between the full cohort and the analysed subset could help readers assess whether excluding patients with missing survival data introduced bias into the dataset.

4. Clarity in Discovery and Validation Cohort Design The rationale for using separate datasets for prognostic (TCGA-SKCM) and predictive (anti-PD-1-treated cohort from Gene Expression Omnibus, GSE91061) analyses requires clearer explanation. It is not explicitly stated whether the validation cohort was independent by necessity (due to treatment data availability) or by design. The paper should clearly justify the selection of each dataset and describe the inclusion/exclusion criteria. A schematic diagram summarising cohort derivation would improve transparency.

5. Reproducibility and Code Availability We commend the author on reusing publicly available datasets. However, the analysis code is only available “upon reasonable request.” For a computational biomarker study, this limits independent verification and reuse. We strongly encourage the author to deposit the full analysis code, documentation, and environment details in a public repository (e.g., GitHub or GitLab).

6. Sample Size and Statistical Power While cohort sizes appear reasonable, the potential impact of limited statistical power was not adequately discussed. For example, the anti-PD-1 cohort has 49 patients and 10 responders. Indeed, two of the metabolic/immune groups only include one responder each. The non-significant survival trends may reflect insufficient power rather than the absence of a biological effect. The wide confidence intervals in regression outputs reflect instability of estimates. We encourage the author to include a detectable-effect/power discussion to help readers contextualize null findings and the interaction analysis.

List of minor concerns and feedback

1. Presentation of Figures and Tables Several tables (3a, 3b, 4) and Figure 3 would benefit from a more detailed description in the Results section to ensure stronger alignment between text and displayed data. Table 4 formatting could be improved for readability. Figure 2 labels overlap with plotted lines, and axis labels lack sufficient detail. Figure 4’s legend requires a clearer explanation of symbols and statistical annotations. Additionally, red-green color schemes may pose challenges for color-blind readers. It is advisable to revise figure layouts, clarify axis labels (including units), expand legends, adjust spacing in tables, and adopt color-blind-friendly palettes.

2. Discussion Depth and Contextualization Comparison with similar transcriptional biomarker studies is limited, and several references are more than five years old. In-text citation formatting is inconsistent, and some references lack corresponding citations. The study would benefit from incorporating recent literature on metabolic-immune interactions in melanoma and refining citation formatting to ensure consistency and completeness.

3. Clarification of Dataset Trimming The rationale for excluding 156 patients is not sufficiently detailed. It is unclear whether the absence of clinical data may itself relate to prognosis or treatment patterns. The authors were advised to provide a flow diagram and discuss possible bias introduced by incomplete clinical annotation.

4. Ethical Transparency Although publicly available datasets were used, a clearer statement confirming adherence to the original study's ethical approvals and data-use policies would strengthen transparency.

5. Limitations Section Expansion The authors appropriately acknowledge several limitations, including bulk RNA-seq constraints and complex metabolic-immune interactions. However, limited discussion is provided regarding unmeasured confounding, comorbidities, and sample size constraints. It is necessary to expand the final paragraphs to explicitly discuss these factors and their potential impact on interpretation.

6. Educational Value of Negative Findings The manuscript commendably reports a negative result without overstating borderline statistical trends. Emphasizing this as a strength, demonstrating rigorous hypothesis testing could enhance the paper’s contribution to methodological transparency.

Overall Assessment

The study rigorously evaluates a biologically plausible hypothesis using large independent cohorts and a structured statistical framework. However, concerns related to cell-type confounding, incomplete covariate adjustment, potential selection bias, and limited reproducibility transparency must be addressed to ensure robust interpretation. Strengthening methodological clarity and expanding sensitivity analyses would substantially enhance the manuscript’s credibility and impact.

Concluding remarks

We thank the author of the preprint for posting their work openly for feedback. We also thank all participants of the Live Review call for their time and for engaging in the lively discussion that generated this review.

Competing interests

Daniela Saderi was a facilitator of this call and one of the organizers. No other competing interests were declared by the reviewers.

Use of Artificial Intelligence (AI)

The authors declare that they did not use generative AI to come up with new ideas for their review.