Institutional Repository [SANDBOX]
Technical University of Crete
EN  |  EL

Search

Browse

My Space

Machine learning methods for the evaluation of biomolecular markers οf pancreatic cancer and its correlation with embryogenesis

Torakis Ioannis

Full record


URI: http://purl.tuc.gr/dl/dias/28FEB459-8ECE-4376-9A6D-F57211F82FB4
Year 2019
Type of Item Diploma Work
License
Details
Bibliographic Citation Ioannis Torakis, "Machine learning methods for the evaluation of biomolecular markers οf pancreatic cancer and its correlation with embryogenesis", Diploma Work, School of Electrical and Computer Engineering, Technical University of Crete, Chania, Greece, 2019 https://doi.org/10.26233/heallink.tuc.83535
Appears in Collections

Summary

Pancreatic cancer is a highly lethal disease, accounting for manydeaths every year. It is considered as one of the most aggressivetypes of cancer, and one of the major problems is the lackof early detection. A patient is diagnosed with pancreatic canceronly in advanced stages, when the possibility of developinga metastases is high. There is no standard procedure to diagnosehigh risk patients, since they remain asymptomatic in thecancer’s early stages. Surgical resection is regarded as the onlypotentially curative treatment, and adjuvant chemotherapy withgemcitabine or S-1, an oral fluoropyrimidine derivative, is givenafter surgery. Therefore, researchers focus on the procedure ofits creation, at a molecular level. There are four major drivergenes for pancreatic cancer: KRAS, CDKN2A, TP53, and SMAD4.KRAS mutation and alterations in CDKN2A are early events inpancreatic tumorigenesis.Recent researches suggest that there is a correlation of somecritical signaling pathways that are activated during pancreaticcancer tumorigenesis with the procedure of embryogenesis.Though, the lack of an analysis that will be able to extract thesegenes involved in the pathways suggested, both in pancreaticcancer patients and embryogenesis samples is crucial. The aimof this thesis is to apply machine learning methods to nd thebiomolecular markers that are deferentially expressed on pancreaticcancer patients and correlate them with markers fromembryogenesis. Since these markers are extracted, we will usethem as classifiers on different machine learning methods, to tryand classify if they refer to patient or healthy subjects.Our thesis contributes a “ 25 gene signature” of biomolecularmarkers which are involved in signaling pathways found inboth embryogenesis and pancreatic carcinogenesis, obtained viafeature extraction and feature selection methods. These markers are used as classifiers for pancreatic cancer classification,and two machine learning classification models are proposed aswell. The classification models achieved high accuracy levels,and we support the notion that our “25 gene signature” in itsentirety can play a classification role in discriminating patientswith pancreatic cancer from healthy controls.

Available Files

Services

Statistics