The Multilayer Network Approach in the Study of Personality Neuroscience
, Hanneke E. Hulst 1, Leon de Bruin 2,3, Gerrit Glas 2, Jeroen J. G. Geurts 1 and Linda Douw 1,4
Lipeng Ning
Round 1
Reviewer 1 Report
This review paper advocates a multilayer network approach that integrates dynamics in behavioral cognitive and emotional characteristics (BCECs) and neural networks in the study of personality neuroscience. To prepare the readers, this paper first presents concise and comprehensive reviews/introductions on the separate study of personality and network neuroscience techniques and their limitations. Then, the multilayer personality neuroscience approach is introduced to overcome the limitations of current research methods. This paper is well organized and nicely written. I think it provides a nice guideline on the future directions in the related research field. The following are some minor comments that are helpful to further improve the quality of this paper.
1) It was mentioned in several places in the paper that PCA and ICA are factor analysis techniques. Strictly speaking, factor analysis (FA) is a different dimension reduction technique with the underlying statistical models and computational algorithms different from PCA and ICA. To be more rigorous, I suggest that "factor analysis techniques" is changed to "statistical approaches" or "dimension reduction techniques".
2) Structural connectivity was only briefly mentioned in Section 3.1. But most discussions were about functional connectivity based on fMRI, EEG or MEG. It will be helpful for the reader to better understand the context by adding more information about how structural connectivity was estimated, e.g. diffusion MRI, and add discussions about if SC is usefully in the study of personality neuroscience.
Author Response
|
Reviewer |
Reviewer Comment |
Revision |
|
1 |
It was mentioned in several places in the paper that PCA and ICA are factor analysis techniques. Strictly speaking, factor analysis (FA) is a different dimension reduction technique with the underlying statistical models and computational algorithms different from PCA and ICA. To be more rigorous, I suggest that "factor analysis techniques" is changed to "statistical approaches" or "dimension reduction techniques". |
We thank this reviewer for the positive remarks and constructive suggestions for improvement. We agree with this comment and have therefore made the following changes:
I. [Line 95]: ‘factor analysis techniques’ changed to ‘statistical approaches. II. [Line 293]: ‘traits’ changed to ‘factors’ III. [Line 295]: ‘factor analysis techniques’ changed to ‘statistical approaches’ IV. [Lines 276-284]: Note to the editor: These lines are a quote and need to be indented or put in quotations.
|
|
1 |
Structural connectivity was only briefly mentioned in Section 3.1. But most discussions were about functional connectivity based on fMRI, EEG or MEG. It will be helpful for the reader to better understand the context by adding more information about how structural connectivity was estimated, e.g. diffusion MRI, and add discussions about if SC is usefully in the study of personality neuroscience. |
We have added further description of measures of structural connectivity and added a study which explores the role of white-matter integrity with regards to the five-factor model of personality. Please see the following changes:
I. [Line 183-193]: Inserted passages “Structural connectivity refers to the anatomical connections between brain regions. Such connections are most often measured using diffusion weighted MRI techniques which characterise white matter pathways, and may map the orientation and integrity of white matter fibre bundles (Yeh et al., 2020). Along with drawing conclusions about structural connectivity between regions, this structural connectivity data can be analysed in terms of its network properties, allowing conclusions to be drawn about, for example, the efficiency of the structural connections throughout the whole brain. Diffusion weighted MRI, as a measure of structural connectivity, has been explored with regards to the five-factor model of personality (Xu and Potenza, 2012). In this study (Xu and Potenza, 2012), white-matter integrity was implicated in mediating neuroticism and openness-to-experience, though in this instance the properties of the structural network as distinct from structural connectivity, were not considered.” II. [Line 197] ‘Resting-state’ changed to ‘Resting state’ III. [Line 259] ‘Resting-state’ changed to ‘Resting state’ IV. [Line 274] ‘Resting-state’ changed to ‘Resting state’ V. [Line 660] Xu et al (2012) citation added. VI. [Line 662]: Yeh et al (2020) citation added.
|
Reviewer 2 Report
In their manuscript the authors explore the potential of employing the multilayer network analysis framework (an established method in network science) in the study of personality neuroscience. They propose the said framework would naturally help incorporate multimodal data relevant for this field. To this goal they provide a brief review of approaches commonly used in personality neuroscience, with an emphasis on network based approaches and subsequently they provide their perspective on how a multilayer network framework would be applied to improve current state of the art.
Consequently, their work is rather an opinion/perspective paper than a review paper. While I have little background on personality neuroscience, I believe this piece of work could have significant impact, even though the authors provide no evidence/preliminary results. The network formalisms (and particularly multilayer networks) are increasingly becoming a popular tool in modeling multimodal data.
Generally, the paper is well structured.
Few minor aspects, I would like to point out:
- Section 3.1: the authors should better harmonize the text, the transition from connectivity -> default mode network -> neuroimaging technologies comes terse. /while I understand the relevance of the DMN and rsfMRI in personality, are there other large scale networks relevant for personality (e.g. salience, attention maybe?). If authors decide to describe DMN they should at least mention other relevant networks.
- Consistent terminology: authors use network science, graph theory (which is ok), then they mix and use 'network theory' (section 3)
- Table 1: undirected network definition => describing edges in this type of network as bidirectional is inaccurate (that would imply network is directed), would rather use 'non-directed' (or describe non-causal connections).
- Section 3.5: what is NEO-PI-R? this abbreviation and related concepts are not introduced.
- Same section: this section needs extensive phrasing revision, e.g. "this made lead' -> may lead? ; 3rd paragraph starting 'This necessitates...' is unclear, do authors mean 'implies'
- Section 4: 2nd paragraph: 'in the expanding...' please rephrase
- same section 6th paragraph: 'is not be able'-> please rephrase
Author Response
|
Reviewer |
Reviewer Comment |
Revision |
|
2 |
Section 3.1: the authors should better harmonize the text, the transition from connectivity -> default mode network -> neuroimaging technologies comes terse. /while I understand the relevance of the DMN and rsfMRI in personality, are there other large-scale networks relevant for personality (e.g. salience, attention maybe?). If authors decide to describe DMN they should at least mention other relevant networks. |
We thank the reviewer for the encouraging words and the useful suggestions on or manuscript. Agreed. We have referred to other large-scale networks, citing the Yeo et al (2011) paper. We have also added a further study that implicates the role of the salience, default mode and executive networks in mediating the FFM trait Openness. There is a vast number of studies out there on this topic, which is slightly beyond the scope of this paper, so they have been added more as examples as opposed to a representative overview. I. [Line 200]: Word inserted ‘functional’ II. [Line 200-204]: Passage inserted: ‘and have allowed for a number of large-scale networks to be identified using resting state functional connectivity data. These include the well-established default mode network, the limbic network and the salience network (Yeo et al., 2011) and have to varying degrees been related to a number of BCECs of personality. For example, the’ III. [Line 216-219] Passage inserted ‘Meanwhile, trait openness-to-experience, which is highly correlated with imagination and intellect, is found to be related to dynamic shifts between a number of large-scale networks including the default mode, salience and executive networks (Beaty et al., 2018). IV. [Line 665]: Yeo et al (2011) citation added. V. [Line 549] Beaty et al (2018) citation added. |
|
2 |
Consistent terminology: authors use network science, graph theory (which is ok), then they mix and use 'network theory' (section 3) |
I. [Line 28]: ‘Network theory’ changed to network science II. [Line 164]: ‘Network theory’ changed to ‘network science’ III. [Line 273]: ‘Network theory’ changed to ‘network techniques’ |
|
2 |
Table 1: undirected network definition => describing edges in this type of network as bidirectional is inaccurate (that would imply network is directed), would rather use 'non-directed' (or describe non-causal connections).
|
Agreed. I. [Table 1, line 7]: ‘Undirected’ network changed to ‘Non-directed’ |
|
2 |
Section 3.5: what is NEO-PI-R? this abbreviation and related concepts are not introduced. |
The NEO-PI-R is introduced on line 98. |
|
2 |
Same section: this section needs extensive phrasing revision, e.g. "this made lead' -> may lead? ; 3rd paragraph starting 'This necessitates...' is unclear, do authors mean 'implies'
|
I. [Line 296]: “this made lead” changed to “this may lead” II. [Line 310] “this necessitates” changed to “this implies” |
|
2 |
Section 4: 2nd paragraph: 'in the expanding...' please rephrase
|
I. [Line 381]: “in the” changed to “by” |
|
2 |
Same section 6th paragraph: 'is not be able'-> please rephrase
|
I. [Line 405]: word removed “be”. |
