Talk:Models of epilepsy
Response to Reviewers
We thank the two reviewers for their very constructive comments. We have made the changes recommended by the two reviewers to the article.
We have added a paragraph about Lytton's work in Section 2.2.
We have added a paragraph about Traub's work in Section 2 (before Section 2.1).
All other minor corrections have been made.
The Scholarpedia manuscripts 'Models of Epilepsy' by Ullah and Schiff is an excellent summary of recent work in the field. I have very few minor comments and suggestions.
- In the third sentence the authors state ' ... irrespective of the age.' Since epilepsy occurs more in younger and older age groups, this could lead to misunderstanding.
- In the section 'Macroscopic models' the authors refer to Lopes da Silva et al (1974). It might be helpful to state explicitly that most other models in this section are cortical models whereas the Lopes da Silva model describes the thalamic component.
- In the section 'Biophysical models' in the 3rd paragraph. The total number of neurons in the model generating the published traces are 656 (512 PCs and 144 INs). The experiments were done in mouse (not in rat) and the animals were younger than 15 days (typically around 10 days). The details about the effect of riluzole on persistent sodium were published in: Drongelen W van, Koch H, Elsen FP, Lee HC, Mrejeru A, Doren E, Marcuccilli CJ, Hereld M, Stevens RL, Ramirez JM. (2006) - The role of persistent sodium current in bursting activity of mouse neocortical networks in vitro. J. Neurophysiol. 96:2564-77.
- I think the article might include a summary of the original work of Bill Lytton and Roger Traub's group.
The article entitled “Models of epilepsy” is a timely review about the main aspects of computational modeling applied to the study of epilepsies. As mentioned in the introduction, the topic is not fully covered here, nor would it have been possible within the present volume of the work. Nevertheless, the authors have managed to represent the major challenges that computational neurosciences face in this field and the most important approaches. The text is overall clearly written and logically organized. References are adequate and to my knowledge contemporary. Therefore, I only have a limited number of comments.
In section 1, among the advantages of macroscopic models, authors mention that their potential usefulness for identification of brain structures that belong to the seizure onset zone. I do agree but I would strengthen this statement. Indeed, in most of the cases, epilepsy can not be reduced to the involvement of a restricted area of the brain. Very often, epileptic activity (either interictal or ictal) spreads over quite extended regions and involves several structures (cortical and sub-cortical). Since they remain relatively simple (size of differential systems), macroscopic models represent the best alternative to physiologically describe epileptic processes occurring in ‘large-scale’ systems. As a minor comment, in figure 1-B, I noticed that an inhibitory connection (from dendritic-projecting INs to somatic-projecting INs) is missing.
In section 2, I think that some references to Traub’s work should be added. Indeed, the detailed approach was extensively developed by this group since the 80s and led to innovative hypotheses about tissue excitability, role of interneurons and factors leading to hypersynchronization (see for instance: Traub RD, Wong RK. Cellular mechanism of neuronal synchronization in epilepsy. Science. 1982 May 14;216(4547):745-7 for a pioneer publication about detailed computational modeling in the field of epilepsy).
When authors mention biophysical models of “networks composed of millions of neurons”, they probably refer to the project called “Blue Brain”. A link to this project aimed at reproducing a neocortical column composed of accurately modeled pyramidal cells interconnected through 30 million synapses with precise 3D locations) could be provided.
In section 3, the opposition between lumped-parameter macroscopic models and detailed network models seems too strong. To me, both approaches are complementary as they can both provide insights about (possibly altered) neuronal systems, at different levels of description. The choice for the modeling approach also strongly depends on the question under study and on the nature of real observations (for instance, electrophysiological data recorded at intra- or extra-cellular level). As mentioned by authors, each approach has its own advantages and limitations. As a perspective, it could be mentioned that “across-scale” approaches could contribute to establish some relationships between sub-cellular/cellular variables of detailed models and “aggregated” parameters governing macroscopic models.
User 4: Clustering coefficient / small world networks
Please reconsider your use of "clustering coefficient" - the term commonly refers to a property of a network (graph) as a whole. The "local clustering coefficient" would refer to the clustering properties of a single node, but even then it is not the number of its connections (as stated in the footnote). The (mean) number of connections of each node is usually referred to as "degree".
The average local clustering coefficient (network c.c.) is relevant w.r.t the small-world property of a network, whereas a HIGH clustering coefficient indicates small-world, not a low one as stated (Watts and Strogatz 1998). Random graphs ("regular graph"?) have lower network clustering coefficients.
A low mean distance is indeed indicative of a small-world network. --Reitter 16:31, 10 February 2010 (EST)
Ullah & Schiff
We made the necessary corrections pointed out. Thank you.