Supervisors info:
Δρ. Γεώργιος Παυλόπουλος - Κύριος Ερευνητής B’ – Ε.ΚΕ.Β.Ε “Αλέξανδρος Φλέμινγκ”
Δρ. Παντελής Μπάγκος - Επιβλέπων ΠΜΣ - Καθηγητής Τμήματος Πληροφορικής με Εφαρμογές στη Βιοϊατρική, Πανεπιστημίου Θεσσαλίας
Δρ. Βασιλική Οικονομίδου - Αναπληρώτρια Καθηγήτρια Βιοφυσικής – Μοριακής Βιοφυσικής,
Τμήματος Βιολογίας, Ε.Κ.Π.Α
Summary:
As clustering of biological networks, it’s called the process according to which the nodes of a network can be classified in a common group according to their common features. In order to succeed this process automatically, there are a variety of clustering algorithms based on different strategies and methodologies. Note that most of these algorithms take into account the network topology. Therefore, different clustering algorithms can often bring out different results even for the same data set.
In order to proceed with this Thesis, initially multiple biological networks of different types were collected from various databases. The collected biological networks that were analyzed are protein interaction networks, gene co-expression networks, and sequence similarity networks. Firstly, these networks were studied in terms of their topology. After that, different data clustering algorithms were applied, the results of which were compared both with each other and with data sets that came after the application of flows related to the functional enrichment of the nodes of each network. Also, as part of this Thesis, it’s presented the conductance measure which shows, through histograms, the quality of each cluster in the rest of the original network. Finally, through the development of the VICTOR tool, various comparison metrics can be applied through which the results of clustering algorithms can be compared via visual analysis.
Keywords:
Networks, Network Clustering Algorithms, Functional Analysis, Conductance
