Review Reports

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Although the paper identifies an interesting correlation between MDD and functional connectivity, it makes no effort in assessing the possibility of assessin MDD in individual people.

I realize that identifying MDD in individual people using functional connectivity is statistically challenging, it would be such an assessment that would greatly increase the novelty and interest of the paper. Including such assessment in the Discussion or in an Attachment would have greatly improved the paper.

The English of the paper and the clarity of the presentation is excellent.

Author Response

Comments 1: Although the paper identifies an interesting correlation between MDD and functional connectivity, it makes no effort in assessing the possibility of assessin MDD in individual people.

I realize that identifying MDD in individual people using functional connectivity is statistically challenging, it would be such an assessment that would greatly increase the novelty and interest of the paper. Including such assessment in the Discussion or in an Attachment would have greatly improved the paper.

The English of the paper and the clarity of the presentation is excellent.

Response 1: Thank you for pointing this out. We agree with your comment. Such individual-level assessment would require dedicated predictive frameworks with out-of-sample validation, careful harmonisation of site effects, and integration of detailed clinical phenotyping, which were beyond the scope of this secondary data analysis plan.

Although the present work was not designed to derive individual-level diagnostic classifiers, the robust alterations in posterior DMN and occipito-parieto-temporal state occupancy identified here delineate candidate dynamic features that future machine-learning and normative-modelling studies could evaluate for subject-level risk stratification and treatment prediction.

The relevance of this limitation, which the esteemed reviewer pointed out, led us to add the following paragraph to the limitations mentioned in the Discussion section of the article (ninth paragraph of the Discussion, page 12, lines 366-375)): "A further limitation is that the current LEiDA analysis was restricted to group-level comparisons between MDD and HC, and was not optimised for individual-level diagnostic classification or prognosis. Developing clinically useful biomarkers would require explicitly predictive modelling with cross-validated training and test sets, harmonisation of site-related variance, and integration of longitudinal and treatment data, which the REST-meta-MDD dataset does not yet provide in a uniform way. Nonetheless, the specific pattern of decreased posterior DMN occupancy and increased occipito-parieto-temporal occupancy identified here offers a mechanistically grounded feature space that can be leveraged by future work targeting subject-level prediction in MDD."

 

 

Reviewer 2 Report

Comments and Suggestions for Authors

Reviewer Comments
General assessment

The manuscript addresses an important topic using a large multicentre dataset and an advanced dynamic functional connectivity approach. However, several methodological and interpretative issues need to be addressed to strengthen the validity, transparency, and interpretability of the results. In particular, limitations related to clinical heterogeneity, medication status, and multicentre acquisition should be more explicitly considered. Please find attached my comments pont-by-point. 

Major Comments
1. Materials and Methods – Clinical and methodological covariates

2.1. Absence of clinical covariates
The study focuses exclusively on group differences between patients with MDD and healthy controls, without considering relevant clinical covariates (e.g., symptom severity, illness duration, episode status). While this choice may be justified by data availability or heterogeneity, this limitation is not sufficiently acknowledged. The Discussion section should explicitly address how the absence of clinical covariates limits the interpretability of the findings.

2.2. Multicentre data and batch effects
Given that data were acquired across multiple cohorts and acquisition sites, it is essential to clarify whether differences between centres were examined or statistically controlled. The manuscript does not specify whether any correction or harmonisation procedure (e.g., site as a covariate, stratified analyses, or batch-effect correction) was applied to minimise potential scanner- or site-related effects. This point should be clarified in the Methods and discussed as a limitation if no correction was implemented.

2.3. Medication status
The sample includes both medicated and unmedicated patients with MDD. However, medication status is neither included in the analyses nor discussed as a potential confound. Given the well-documented effects of psychotropic medication on resting-state functional connectivity, the authors should justify this omission and explicitly discuss its possible impact on the reported results.

2.4. Scanner field strength (1.5 T vs 3 T)
The analysis does not consider potential differences between data acquired at 1.5 T and 3 T. Field strength can substantially affect signal-to-noise ratio and functional connectivity estimates. At a minimum, this issue should be acknowledged and discussed as a limitation in the Discussion section, and the authors should clarify whether scanner type was balanced across groups.

2.6. Definition of phase-locking states
In the Methods section, the authors should explicitly define what they consider a “phase-locking” state, including the theoretical and numerical limits of the measure (e.g., −1 to +1). This clarification is necessary to avoid confusion with other commonly used metrics such as the phase-locking value (PLV) employed in oscillatory analyses, which may be conceptually conflated with the present approach.

Data analysis: 

The authors explore a wide range of clustering solutions (K = 2–20) and subsequently highlight two specific states (K18C4 and K20C18) as the main findings. Although a Bonferroni correction is applied, the manuscript does not clearly describe any a priori criterion for selecting either the optimal K or the specific states deemed of interest. This analytical strategy introduces a substantial risk of post hoc selection and inflates the likelihood of reporting chance findings, even in the presence of multiple-comparison correction. The authors should explicitly define an a priori criterion for K selection (e.g., based on clustering stability, reproducibility across subsamples or sites, or cross-validation metrics). Then, please clearly distinguish between Confirmatory analyses, based on pre-registered or objectively defined K values, and Exploratory analyses, covering the remaining range of K values. Without this distinction, the current narrative gives the impression that results were selected retrospectively, which undermines the strength of the statistical inference.

In addition, the description of multiple-comparison correction is ambiguous. The manuscript refers both to a Bonferroni correction across 209 states (corresponding to the total number of clusters across K = 2–20) and to family-wise error (FWE) correction within each K. These are conceptually and statistically different approaches. The authors must clearly specify which correction strategy is ultimately used to support the main conclusions,at which inferential level (global vs. within-K), and ensure that the reported p-values and significance claims are fully consistent with the chosen correction scheme.

Robustness and LEiDA preprocessing pipeline: there is an inconsistency in the description of temporal preprocessing and filtering. In the general preprocessing section, the authors report applying a band-pass filter (0.01–0.08 Hz). However, in the description of the LEiDA pipeline, it is later stated that time series were detrended and demeaned “without additional filtering” prior to Hilbert transform. Did they use band-pass filter or not. Given that phase estimation via the Hilbert transform is highly sensitive to filtering choices, this point must be clarified explicitly.

Minor Comments
Abstract

Please revise the punctuation of Hedges’ g throughout the Abstract. In at least one instance it appears as Hedge´s, which is inconsistent and incorrect.

Data visualisation

Figures 4, 5, 6, and 7: The current boxplot visualisations are suboptimal. A large number of data points appear above the upper quartile, with virtually none below the lower quartile, which raises concerns about interpretability and distributional transparency. Alternative visualisations, such as violin plots or scatter plots showing all individual data points (possibly overlaid with summary statistics), would provide a more informative representation of the data distributions.

Discussion

The authors should explicitly state that differences between clinical subtypes of MDD were not examined and discuss how this omission may limit the generalisability of the findings.

Given the multicentre nature of the dataset, the Discussion should also address the robustness of the results across acquisition centres and scanner types (1.5 T vs 3 T), particularly if no formal correction or stratified analysis was performed.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The Authors respond well to all my concerns. However, despite they state that the quality of figures 4 tp 7, the visualization of the statistical data with boxplots can still be improved (maybe with different axis scales and width). Also, in the present manuscript, figures 2 to 3 have loosen definition. 

Author Response

Comment 1: "The Authors respond well to all my concerns. However, despite they state that the quality of figures 4 tp 7, the visualization of the statistical data with boxplots can still be improved (maybe with different axis scales and width). Also, in the present manuscript, figures 2 to 3 have loosen definition."

Response 1: We sincerely thank you for your valuable feedback, which has strengthened the manuscript. We improved the quality of figures 2 to 7, has suggested. Additionally, we also streamlined the Title to "Dynamical Exploration of Brain Attractors at Rest Altered in Major Depressive Disorder" for precision and impact.