University of Patras
Μέθοδοι μηχανικής εκμάθησης από δεδομένα μεγάλης κλίμακας με εφαρμογή στην ανάλυση δικτύων της μοριακής βιολογίας και των νευροεπιστημών
Abstract
dc:descriptionIn this PhD Dissertation, the aim was creation and analysis of biological networks using Machine Learning methods. Towards this direction, new methodologies were developed for modelling, reconstruction and analysis of networks, providing valid biological results. Different techniques, such as clustering, classification, regression and feature selection were successfully combined with graph theoretical analysis to study different kinds of biological networks. Specifically, the examined networks varied from microscopic scale at the level of the cell to macroscopic scale at the level of the brain, derived from Molecular Biology and Neuroscience data, respectively.Systems Biology studies the cellular mechanisms as a system, using as a key tool the graph theory on heterogeneous experimental data of Molecular Biology. In Molecular Biology networks, genes, RNA molecules and proteins are considered as nodes for modelling the complex interaction of these entities and for analyzing their behavior in complex diseases and conditions, such as cancer types and aging. In detail, large scale data measuring gene expression (microarrays) were used either for building networks de novo or in combination with existing networks, such as gene regulatory networks, protein interaction networks and cellular pathways. Specifically, network modelling methods were developed using regression random forests aiming to predict new interactions or to detect perturbations in the network. Additionally, as a result of network analysis, sets of differentiated nodes were detected, which can be utilized for understanding the disease effects and additionally can be potential drug targets.In Neuroscience field, electroencephalography experiments were carried out with high sensor density in order to study high order cognitive functions, such as mental workload and mental fatigue. From these data, networks were reconstructed representing the interactions among brain areas in frequency domain, which were analyzed with graph theoretical metrics. Additionally, using these weighted graphs as features for selection and classification, a small subnetwork of important interactions was isolated, which achieved high accuracy in discriminating between different conditions. This subnetwork can contribute to comprehension of cognitive mechanisms and enables efficient monitoring of brain activity.
Degree
thesis:*- Grantor dc:publisher
- University of Patras
- Year dc:date
- 2018
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Dimitrakopoulos, Georgios
- Δημητρακόπουλος, Γεώργιος
Subjects
dc:subject × 16- Μηχανική εκμάθηση
- Γράφοι
- Συστημική βιολογία
- Υπολογιστικές νευροεπιστήμες
- Machine learning
- Graphs
- Systems biology
- Computational neuroscience
- Φυσικές Επιστήμες
- Επιστήμη Ηλεκτρονικών Υπολογιστών και Πληροφορική
- Επιστήμες Μηχανικού και Τεχνολογία
- Βιοϊατρική Μηχανική
- Natural Sciences
- Computer and Information Sciences
- Engineering and Technology
- Medical Engineering
Rights
dc:rights- Statement dc:rights
-
- BY_NC
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/43575