Understanding of Depressive Symptomatology across Major Depressive Disorder and Bipolar Disorder: A Network Analysis

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript by Lee et co-authors describes the analysis of a sample of self-rating depression severity scales in patients with MDD and BD.

Relatively balanced dataset, which is enough for multivariate comparisons, was collected with a help of the Zung Self-Rating Depression Scale.

This dataset was screened by the exploratory graph analysis for possible clusters of symptoms, without further necessary and sufficient multivariate analyses employed.

It is possible to increase the value of the data reported in the following ways.

 

[1]

The exploratory graph analysis (EGA) is a good approach to estimate a number of dimensions for any specific multivariate dataset, which must be accounted for to dissect intervening pathologies.

However, comparing network structures of self-rating depression severity scales in patients with MDD and BD is not sufficient to elucidate the central depressive symptoms in both groups and explore connections among clustering of symptoms.

After EGA, the analysis of data must be extended by more traditional multivariate techniques including PCA and non-metric dimension reduction techniques like PacMAP, UMAP or tSNE.

 

[1a]

The Fig.1 depicting the network structure is nice and traditional for the field, yet visualizes only association levels between individual estimates.

To strengthen the conclusions, the authors must do the classic PCA, visualizing the scree plot, and critical PCxPC planes, presenting the localisation of variables in the space of principal components impacting the variability.

 

[1b]

The authors also must perform the non-metric dimension reduction technique like UMAP, PacMAP or resource-expensive tSNE to estimate the number of separable clusters independently of EGA.

The results of a UMAP (PaCMAP, tSNE) dimension-reduction projection must be visualized.

To achieve this, please see Wang et al. (2021) Understanding how dimension reduction tools work: an empirical approach to deciphering t-SNE, UMAP, TriMap, and PaCMAP for data visualization. Journal of Machine Learning Research 22 (2021) 1-73

https://dl.acm.org/doi/abs/10.5555/3546258.3546459

 

[1c]

The results of the PCA, UMAP (PaCMAP, tSNE) and EGA must be compared and discussed in terms of identified separable clusters.

 

[2]

Demographic and clinical characteristics must be visualized as Raincloud plots with jittering, the Median, IQR boxes and CI95 intervals visualized for each quantitative characteristic like age, between control and affected samples.

 

[3]

Finally, the authors might test the proposed variables as predictors of specific pathological states, using Random forest, boosting or other ML-based techniques of classification.

In such case, it would be valuable to depict the ROC curves for a dataset classified.

Comments on the Quality of English Language

Some clauses are fuzzy and must be simlified.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

 

1)     Manuscript language needs improvement as there are many formatting, grammar, spelling, spacing issues etc.

2)     In abstract line 25 four distinct communities’ word is used which should be four categories

3)     In line 41 and 42 the recurrence rate mentioned in introduction does not match with percentage mentioned in Reference 8 and 9.

4)     Results 3.1 line 163 says that the cohort was predominantly female (72.2 %).will it not affect the results? As alcohol use is considered in cohort. And alcohol is used more by males rather than females.

Comments for author File: Comments.pdf

Comments on the Quality of English Language

1)     Manuscript language needs improvement as there are many formatting, grammar, spelling, spacing issues etc.

2

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

[1]

The authors done a great work completing the separate PCA analyses for MDD and BD participants.

The exploratory PCA confirms the high similarity of general clusters identified for participants with distinct disorders by Exploratory Graph Analysis, revealing that EGA is a possible approach for estimating the number of dimensions in psychological research.

In addition, the exploratory PCA evidences that the EGA might misclassify some of individual questionary items between clusters identified, which are classified routinely by PCA in accordance to their similarity with PCs identified.

This finding suggests that EGA might not be used as a standalone technique to uncover the structure of data in psychiatric research, suggesting EGA must be performed in pair with PCA for proper interpretations of data-underlying associations.

Please indicate this critical finding both in the Results section, in the Limitation subsection of the Discussion section and in the Conclusion section of the manuscript.

Please also consider to reflect this critical finding in the Abstract (see the point of 4 this review )

 

[2]

The authors suggested that “The network comparison test did not reveal significant differences in network structure between MDD and BD”. Nevertheless, the PCA analyses revealed similarities and differences of MDD and BD participants in terms of latent variables.

In particular, the PC1s, PC2s and MDD’s PC4-BD’s PC3 are similar in MDD and BD participants, while MDD’s PC3-BD’s PC4 might be considered as a potential basis for discrimination between BD and MDD.

This finding is coherent with the specificity of MDD and BD disorders, in which MDD disorders affect the quality of life in people more seriously.

This finding further suggests that EGA might not be used as a standalone technique to uncover the structure of data in psychological research and must be applied in pair with PCA for proper interpretations of data-underlying associations.

[2a]

Please reflect the findings on MDD and BD PC’s associations in the separate Results section on PCA.

[2b]

In the future work, the reviewer also recommends performing the Canonical analysis with both MDD and BD participants to identify differences between these groups.

 

[3]

As a derivate, the PCA results mentioned in [1] and [2] of this review provide a basis on the visual representation of Network structures.

In the present form, the Network structure depicted on the Panels 1A and 1B of the Fig.1 is drawn without any context on X and Y dimensions.

Please consider to revise the visual representation of Network nodes on the Panels 1A and 1B, pinning them to Factor1-Factor2 X-Y positions identified by the PCA.

 

[4]

In the present form, the “Conclusions” section of the Abstract replicates its “Results” section.

Please improve.

 

 

Minor issues

[5]

In the Limitations subsection of the Discussion section, please note that differences between MDD and BD might be due to age of patients.

 

 

Comments on the Quality of English Language

[6]

Please correct the typos for Group 4 in the panels of Fig. 1.

 

Author Response

"Please see the attachment."

Author Response File: Author Response.pdf