Poboljšanje sastavljenih genoma metodama dubokog učenja

master's thesis

• Title (english): Assembly Improvement Using Deep Learning Methods
• Author: Antonio Jurić
• Mentor: Mile Šikić (mentor)
• Committee member: Mile Šikić (predsjednik povjerenstva)
• Committee member: Marko Subašić (član povjerenstva)
• Committee member: Krešimir Križanović (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2018-09-20, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Razvijanje alata koji sastavljaju genom s dovoljnom točnošću da bi bili upotrebljivi
u praksi još je otvoren problem. Uređaji treće generacije za sekvenciranje genoma
omogućuju sastavljanje genoma s manjom stopom fragmentiranosti te visokom točnošću.
Konačnu točnost sastavljenog genoma u ovom radu pokušavamo popraviti metodama
dubokog učenja. Trenirana konvolucijska duboka mreža na temelju naučenih frekvencija
pojedinih baza, umetanja i brisanja popravlja konačni genom ispravljajući pronađene pogreške. Mreže su pripremljene za Pacific Biosciences te Oxford Nanopore
podatke. Trenutni rezultati pokazuju da modeli na nekim konsenzusima neznatno
poprave, dok na drugima neznatno snize druge točnost. No, modele je moguće još
popraviti s nizom neisprobanih ideja opisanih na kraju rada.

Abstract (english)

Development of tools for genome assembling with enough precision to be useful
in practice is still an open problem. Third generation sequencers allow genome assembly
with smaller fragmentation and high accuracy. In this work, we try to improve
the final accuracy of the assembled genome using deep learning techniques. Trained
convolutional deep network polishes errors in the assembled genome by learning the
correct bases, insertions and deletions patterns. Networks are prepared to be used with
Pacific Biosciences and Oxford Nanopore data. Current results show that sometimes
models slightly improve, sometimes slightly degrade consensus quality, but there is
still a room for progress since there are a lot of untested ideas which are described in
the end of the thesis.

• Keywords: bioinfromatika; konsenzus; sastavljanje genoma; duboko učenje
• Keywords (english): bioinformatics; consensus; genom assembly; deep learning
• Language: english
• URN:NBN: urn:nbn:hr:168:912275
• Study programme: Title: Computing; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2019-03-13 21:19:50