1. Αποθετήριο Παραδοτέων Υποτρόφων Δράσεων ΕΣΠΑ Τριτοβάθμιας Εκπαίδευσης 2017-2023
2. Όλο το Αποθετήριο
3. Ερευνητικές ομάδες

Τύπος:	Αναρτημένη ανακοίνωση (poster)
Τίτλος:	Methylation signatures with diagnostic value in breast cancer via automated machine learning
Εναλλακτικός τίτλος:	Βιοϋπογραφές μεθυλίωσης με διαγνωστική αξία στον καρκίνο του μαστού
Συγγραφέας:	[EL] Παναγοπούλου - Πανταζή, Μαρία[EN] Panagopoulou - Pantazi, Maria [EL] Καραγλάνη, Μακρίνα[EN] Karaglani, Makrina [EL] Μανωλόπουλος, Ευάγγελος[EN] Manolopoulos, Vangelis
Επικεφαλής ερευνητικής ομάδας:	[EL] Χατζάκη, Αικατερίνη[EN] Chatzaki, Ekaterini
Ημερομηνία:	17/03/2021
Περίληψη:	Goals: DNA methylation is a well-characterized epigenetic mechanism that plays a key role in the pathophysiology of Breast Cancer (BrCa). Currently, genome wide approaches have broadened our knowledge about DNA methylation and Machine Learning (ML) offers a powerful alternative in analyzing these high-throughput data. Our task was to employ automated ML (autoML) technology on genomewide BrCa methylation datasets in order to produce accurate diagnostic signatures of clinical value in BrCa personalized management. Methods: Publicly available high-throughput methylation raw data from BrCa and normal breast tissues were retrieved by TCGA and GEO. Data from 490 Primary BrCa, 30 metastatic BrCa, and 185 normal tissues were subjected to RnBeads, and normalized gene β-values were used for further ML analysis. Just Add Data Bio (JADBio, www. jadbio.com/), an innovative autoML tool, was applied to build diagnostic signatures. Extensive tuning effort was used and sample datasets were automatically split into training and validation groups in a proportion of 70/30. Results: Multiclass classification analysis of the training set produced a best performing five-gene signature by Support Vector Machines that was able to discriminate among primary BrCa, metastatic BrCa and healthy breast tissue with an AUC of 0.991 [0.973, 0.999]. Validation showed an AUC of 0.998 verifying the model’s performance stability. A best interpretable model of a five-gene signature was also built using Classification Decision Tree algorithm, with an AUC of 0.938 [0.884, 0.985]. Its validation showed an AUC of 0.956. Conclusion(s): Methylation profiling holds valuable information for breast cancer diagnosis. Multiclass classification analysis produced accurate and low feature signatures of stable performance with great translatability to multiplex assays. Genes selected are subjects of further research in breast cancer biology. Our results highlight the potential of autoML analysis of –omics datasets for driving clinically relevant solutions. This research is carried out and funded in the context of the project “DNA methylation as a minimally-invasive biomarker: development and validation of classifiers with prognostic and/or predictive clinical value in breast cancer therapy” (MIS 5049913) under the call for proposals EDULLL 103. The project is co-financed by Greece and the European Union (European Social Fund- ESF) by the Operational Programme Human Resources Development, Education and Lifelong Learning 2014–2020.]
Γλώσσα:	Αγγλικά
Τόπος δημοσίευσης:	Vienna, Austria
Σελίδες:	1
DOI:	10.1016/S0960-9776(21)00137-5
EISSN:	1532-3080
Θεματική κατηγορία:	[EL] Βασική ιατρική, άλλοι τομείς[EN] Basic medicine, miscellaneous
Λέξεις-κλειδιά:	Breast Cancer Diagnosis; machine learning; Epigenetics
Κάτοχος πνευματικών δικαιωμάτων:	DUTH © 2021 The Author(s) © 2021 Elsevier Ltd. Published by Elsevier Ltd.
Ηλεκτρονική διεύθυνση του τεκμηρίου στον εκδότη:	https://www.sciencedirect.com/science/article/pii/S0960977621001375
Ηλεκτρονική διεύθυνση περιοδικού:	https://www.sciencedirect.com/journal/the-breast
Τίτλος πηγής δημοσίευσης:	The Breast
Τόμος:	56
Μέρος:	Supplement 1
Σελίδες τεκμηρίου (στην πηγή):	S38
Όνομα εκδήλωσης:	17th St. Gallen International Breast Cancer Conference
Τοποθεσία εκδήλωσης:	Vienna, Austria & Online
Ημ/νία έναρξης εκδήλωσης:	17/03/2021
Ημ/νία λήξης εκδήλωσης:	20/03/2021
Σημειώσεις:	This research is carried out and funded in the context of the project “DNA methylation as a minimally-invasive biomarker: development and validation of classifiers with prognostic and/or predictive clinical value in breast cancer therapy” (MIS 5049913) under the call for proposals EDULLL 103. The project is co-financed by Greece and the European Union (European Social Fund- ESF) by the Operational Programme Human Resources Development, Education and Lifelong Learning 2014–2020.]
Εμφανίζεται στις συλλογές:	Ερευνητικές ομάδες