Full text
| |
| Author(s): |
Lucas Antiqueira
Total Authors: 1
|
| Document type: | Doctoral Thesis |
| Press: | São Carlos. |
| Institution: | Universidade de São Paulo (USP). Instituto de Física de São Carlos (IFSC/BT) |
| Defense date: | 2011-12-13 |
| Examining board members: |
Luciano da Fontoura Costa;
Alexandre Gonçalves Evsukoff;
Nina Sumiko Tomita Hirata;
Elbert Einstein Nehrer Macau;
Igor Polikarpov
|
| Advisor: | Luciano da Fontoura Costa |
| Abstract | |
The relationship between structure and dynamics was addressed by employing a wide range of different approaches. First, the correlations between degree and activity were studied in various real-world networks. The activity is defined as the proportion of visits to each node in the steady-state regime of the simple random walk. This type of correlation can provide means to assess node activity only in terms of the degree. The concept of accessibility was included in this analysis, showing an intimate relationship (in networks such as the WWW) between the type of correlation and the level of accessibility observed on nodes. A new complex network model founded on growth was also proposed, with new connections being established proportionally to the current activity of each node. This model can be understood as a generalization of the Barabási-Albert model for directed networks. By using several topological measurements we showed that this new model provides, among several other traditional theoretical types of networks, the greatest compatibility with three real-world cortical networks. Additionally, we developed a novel approach considering non-overlapping subgraphs and their interrelationships and distribution through a given network. The main aspect of the methodology is a novel merging procedure developed to assess the relevance of nodes (in relation to the overall subgraph interconnectivity) lying outside subgraphs. Experiments were carried out on four types of network models and five instances of real-world networks, in order to illustrate the application of the method. Furthermore, these results were related to the properties of the transport and spreading processes. Other topic here addressed is the sampling problem in cortical networks. Effects of sampling were quantified using multivariate analysis and classifiers based on structural network measurements. Samples were also evaluated in terms of their dynamical behavior using a synchronization model and the measure of accessibility. By simulating MEG/EEG recordings it was found that sampled networks may substantially deviate from the respective original networks, mainly for small sample sizes. We also report an analysis of the integrated network of Escherichia coli, which incorporates (i) transcriptional regulatory interactions, (ii) metabolic/signaling feedback and (iii) protein-protein interactions. Network outliers, which represent global transcriptional regulators, were identified in the relationship between out-degree and activity. These outliers are highly and widely expressed across conditions, therefore supporting their global nature in controlling many genes in the cell. (AU) | |