Συντάχθηκε 02-10-2013 08:45
από Eleni Stamataki
Email συντάκτη: estamataki<στο>tuc.gr
Ενημερώθηκε:
-
Ιδιότητα: σύνταξη/αποχώρηση υπάλληλος.
Σχολή Ηλεκτρονικών Μηχανικών και Μηχανικών Υπολογιστών
Πρόγραμμα Προπτυχιακών Σπουδών
ΠΑΡΟΥΣΙΑΣΗ ΔΙΠΛΩΜΑΤΙΚΗΣ ΕΡΓΑΣΙΑΣ
ΝΙΚΟΛΑΟΥ-ΚΟΣΜΑ ΧΛΗ
με θέμα
Σύγκριση Στατιστικών Μεθόδων Εξαγωγής Γονιδιακών Υπογραφών
Comparison of Statistical Methods for Genomic Signature Extraction
Παρασκευή 4 Οκτωβρίου 2013, 3μμ
Αίθουσα 145.Π58, Κτίριο Επιστημών, Πολυτεχνειούπολη
Εξεταστική Επιτροπή
Καθηγητής Μιχάλης Ζερβάκης (επιβλέπων)
Αναπληρωτής Καθηγητής Κώστας Μπάλας
Καθηγητής Απόστολος Δόλλας
Abstract
In recent years microarray technologies have gained a lot of popularity for their ability to quickly measure the expression of thousands of genes and provide valuable information for linking complex diseases such as cancer to their genetic underpinnings. Feature selection methods are used in order to extract small and informative sets of genes that can maximize the performance of classification methods used to map unknown samples into classes of interest, leading to new and efficient methods for prognosis of several diseases, which are personalized to the genome of each specific patient. Moreover, the biological interpretation of these sets of genes, often referred to as “genomic signatures”, can help biologists and physicians better understand the biological processes related to complex diseases, such as cancer and may potentially lead to the discovery of new methods of treatment.
Nevertheless, the large number of parameters to be estimated in relation to the small number of available samples gives rise to an “ill posed” problem where the performance assessment of feature selection and classification methods is not stable under slight changes the dataset. In this thesis, a generic evaluation framework named “Stable Bootstrap Validation” (SBV) is presented, that utilizes resampling of the original dataset and an explicit criterion that determines the stability of the observed classification accuracy, as well as the genomic signature. The proposed methodology works in an iterative manner and converges to a stable solution that combines good accuracy with biologically meaningful feature selection. The methodology is orthogonal to the specific feature selection and classification algorithms used. Moreover, methodologies for assessing the statistical significance and consistency of the observed results are also introduced. Some of the most widely used classifiers are compared, based on their average discrimination power and the size of the derived gene signature. According to our proposed model, a unified ‘77 common-gene signature’ was selected, which is closely associated with several aspects of breast tumorigenesis and progression, as well as patient-specific molecular and clinical characteristics.
Συνημμένα:
-
ΑΝΑΚΟΙΝΩΣΗ ΧΛΗΣ 2.docx
Μεταφορτώσεις: 104,
Μέγεθος: 27 KB application/vnd.openxmlformats-officedocument.wordprocessingml.document
