Description
Date depot: 1 janvier 1900
Titre: Integrating information from genetic and neuroimaging data through network approaches
Directeur de thèse:
Fabrizio DE VICO FALLANI (ICM)
Domaine scientifique: Sciences et technologies de l'information et de la communication
Thématique CNRS : Non defini
Resumé:
The traditional way to study complex systems adopted a reductionist approach, where each component is analyzed separately; for instance in molecular biology (or in neuroscience) the role of each gene (or brain area) was studied independently on the others, completely neglecting the possibly existing interactions. By including the role of interactions between components, complex networks theory has established itself as a powerful approach to describe biological complex systems from an holistic perspective. Mathematically, any interconnected system can be modeled by a network considering distinct elements represented as nodes (or vertices) and connections between elements as links (or edges).
Interestingly, the network structure does contain information that can be used to tell if and how the system can be theoretically controlled. The idea is to capitalize on well known results from control theory (e.g. Kalman controllability criterion) that allow to identify driver nodes in interconnected dynamical systems. Recent studies have already started to establish powerful links between network topology and controllability, allowing, for example, the identification of the number of driver nodes (a NP-hard problem) in a significantly reduced amount of time.
This project intends to develop {{novel frameworks combining networks and control theory to study the controllability of brain and molecular networks}}. To fully exploit the information that can be extracted from different biological networks (e.g. different subset of driver nodes in brain and molecular networks), the project eventually intends to develop {{machine-learning techniques to fuse different network features in an effort to improve the classification of diseased states}}. In particular, to combine information from different networks, we will investigate different approaches including multiple kernel learning and multi-task learning.
Doctorant.e: Bassignana Giulia