Model dubokog učenja za određivanje očitanih baza dobivenih uređajem za sekvenciranje MinION

master's thesis

• Title (english): Deep Learning Model for Base Calling of MinION Nanopore Reads
• Author: Marko Ratković
• 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: 2017-07-10, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Uređaji za sekvenciranje MinION tvrtke Oxford Nanopore Technologies su pristupačni i prenosivi što ih čini pogodnim za razne primjene.
Uređaj omogućuje sekvenciranje očitanja velikih duljina ali većeg postotka greške u odnosu na prethodne tehnologije.
Cilj ovog diplomskog rada je pokazati da trenutna pogreška nije uzrokovana isključivo metodom sekvenciranja, već i programskim alatima koji se koriste za očitavanje baza te je pogrešku moguće smanjiti korištenjem metoda dubokog učenja.

Predstavljen je novi alat za očitavanje baza temeljen na konvolucijskim neuronski mrežama koji pruža napredak u preciznosti i brzini u odnosu na trenutno korištene rekurzivne neuronske mrežama. U radu je dana detaljna analiza razvijenog alata i usporedba s postojećim rješenjima za određivanje očitanih baza.

Abstract (english)

MinION by Oxford Nanopore Technologie is affordable and portable sequencing device suitable for various applications. The device produces very long reads, however, it suffers from high sequencing error rate.
The goal of this thesis is to show that the reported accuracy of the sequencing data is not only limited by sequencing technology, but also by the current software tools used for base calling and can be further improved by using different deep learning concepts.
Approach for base calling of raw data using convolutional neural networks is proposed as an alternative to recurrent neural networks used by other basecallers offering improvements both in speed and accuracy. A detailed comparison of the developed tool with the existing tools for base calling R9 data is given.

• Keywords: određivanje baza; Oxford Nanopore Technologies; MinION; duboko učenje; prevođenje; konvolucijske neuronske mreže; rezidualne mreže; CTC gubitak
• Keywords (english): base calling; Oxford Nanopore Technologies; MinION; deep learning; seq2seq; convolutional neural network; residual network; CTC loss
• Language: english
• URN:NBN: urn:nbn:hr:168:308213
• 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: 2019-03-14 00:23:03