Patient Similarity Networks for Irritable Bowel Syndrome: Revisiting Brain Morphometry and Cognitive Features

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

 

This paper focuses on a crucial and relevant issue in the area of gut-brain interaction disorders, specifically the significant variability of Irritable Bowel Syndrome (IBS) and the shortcomings of symptom-centered diagnostic models in reflecting the underlying neurobiological differences. The writers utilize Patient Similarity Networks (PSN) on brain morphometric and cognitive information to investigate data-driven patient subgroups that go beyond conventional diagnostic limits.

The main innovation of the research is the utilization of an unsupervised network-based method on a thoroughly defined IBS cohort featuring multimodal phenotyping. Previous research from the same group utilized supervised machine learning to differentiate IBS patients from healthy individuals, whereas this study takes an additional approach by consciously omitting diagnostic labels in the model development process. The transition from classification to structure discovery signifies an important conceptual progress and aligns closely with contemporary trends in precision medicine and network medicine.

The combination of high-resolution brain morphometry (volumetrics from FreeSurfer) with standardized cognitive assessments (RBANS) significantly boosts the significance of the study. The manuscript effectively shows that patient similarity founded on objective neurobiological and cognitive traits does not align perfectly with the IBS/healthy control division, emphasizing considerable overlap at the individual level. This discovery has clinical significance, as it questions the unspoken belief that symptom based categories must align with separate neurobiological phenotypes.

 
Major Comments

1. Clarification of Methodological Choices in PSN Construction

The methodological framework is robust and well described; however, several analytical decisions would benefit from additional clarification to further enhance transparency and reproducibility.

• The authors describe the use of mean imputation for missing data but do not specify the proportion or distribution of missing values across features. Providing this information, along with a brief justification for the choice of mean imputation, would strengthen the methodological reporting and help readers assess potential impacts on similarity structure.

• The selection of the kernel bandwidth parameter (σ) as the mean pairwise distance is reasonable, but a brief rationale or supporting citation would improve clarity, given the sensitivity of Gaussian kernels to bandwidth choice.

• While the sparsification parameters (k = 8, similarity threshold = 0.3) are supported by sensitivity analyses, it would be helpful to explicitly state whether these parameters were defined a priori rather than optimized post hoc.

These clarifications would not require additional analyses but would meaningfully improve methodological transparency.

 

Minor comments

^{1. External Validation and Generalizability}

While the authors clearly discuss limitations regarding sample size and single-site recruitment, they might want to explicitly emphasize the lack of external validation for the PSN-derived communities. A concise remark highlighting the importance of replication in separate cohorts would enhance the limitations section and conform to best practices in network-based and machine learning studies.

_{2. Scope of Biological Modalities}

The PSN is constructed exclusively from brain morphometry and cognitive measures. While this focus is appropriate, the authors could briefly acknowledge that other relevant domains of the gut–brain axis (e.g., microbiome, inflammatory markers, autonomic measures) are not captured in the current model. This clarification would help contextualize the scope of the findings without detracting from the study’s strengths.

^{Overall Summary}

This is a thoroughly planned, methodologically sound, and carefully analyzed study that offers a significant contribution to the research on IBS heterogeneity and network-oriented patient modeling. The manuscript is appropriate for publication following minor clarifications, especially in the Methods section, and represents a robust instance of exploratory, data-driven research consistent with precision medicine principles.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This study analyzes brain morphometry and cognitive functions using Patient Similarity Networks  to understand the heterogeneity of Irritable Bowel Syndrome.

Overall, the study is methodologically sound, the writing is clear, and it is familiar with the literature. However, the following points are suggested for correction:

1- Although the Louvain algorithm generated 4 communities, scenarios with fewer (k=2–3) or more (k=5–6) communities are not discussed in detail.

2- The discussion should more clearly emphasize that PSN captures a "brain-cognitive dimension independent of symptom severity."

3- It is suggested that the text reframe the communities as "hypothesis-generating neurobiological patterns" rather than "subtypes." The term "phenotype" may be perceived as too strong.

4- The terms "community" and "subgroup" are sometimes used interchangeably; more consistency could be achieved.

5- Figures 1 and 4 are informative but dense. A short "take-home message" box in Figure 4 might be helpful for the reader.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf