Fázování variant ze sekvenačních dat (short reads, Illumina)

Abstract

Haplotype phasing determines the relative arrangement of alleles on ho- mologous chromosomes and is therefore an important step in genetic analysis. In this thesis, we propose an algorithm for haplotype phasing using short-read sequencing data that combines a graph representation of the problem with dynamic programming. The method was evaluated on real human and sim- ulated data. The results showed that the algorithm phases most phaseable variants and achieves a low error rate on simulated data. As the proportion of repetitive sequences increases, the number of phased variants decreases, but without a substantial increase in error rate. The main limitations arise from both the nature of the data, since short reads connect only nearby vari- ants, and the algorithm design, which operates within a fixed local context. The time and memory complexity of the program is O (︂ R · t2 + n · 2k · k2 )︂ , where R is the number of reads, t the maximum number of variants covered by a single read, n the number of variants, and k the window size parameter.

Fa ́zova ́nı ́ haplotypu ̊ umoz ̌n ̌uje urc ̌it vza ́jemne ́ uspor ̌a ́da ́nı ́ alel na ho- molognı ́ch chromozomech, a pr ̌edstavuje tak du ̊lez ̌ity ́ krok v analy ́ze gene- ticke ́ informace. V te ́to pra ́ci navrhujeme algoritmus pro fa ́zova ́nı ́ ze sekve- nac ̌nı ́ch dat s kra ́tky ́mi c ̌tenı ́mi, zaloz ̌eny ́ na grafove ́ reprezentaci proble ́mu a dynamicke ́m programova ́nı ́. Metoda byla analyzova ́na na skutec ̌ny ́ch lid- sky ́ch i simulovany ́ch datech. Vy 'sledky uka ́zaly, z ̌e algoritmus doka ́z ̌e zfa ́zovat pr ̌eva ́z ̌nou c ̌a 'st fa ́zovatelny ́ch variant a na simulovany ́ch datech dosahuje nı ́zke ́ chybovosti. S rostoucı ́m podı ́lem repetic v sekvenci poc ̌et zfa ́zovany ́ch variant klesa ́, avs ̌ak bez vy ́razne ́ho na ́ru ̊stu chybovosti. Hlavnı ́ omezenı ́ majı ́ dvojı ́ pu ̊vod: dany ́ charakterem dat (kra ́tka ́ c ̌tenı ́ propojujı ́ pouze blı ́zke ́ vari- anty) a dany ́ na ́vrhem algoritmu, ktery ́ pracuje v loka ́lnı ́m kontextu pevne ́ho rozsahu. C ̌asova ́ a pame ̌t 'ova ́ sloz ̌itost programu je O (︂ R · t2 + n · 2k · k2 )︂ , kde R pr ̌edstavuje poc ̌et c ̌tenı ́, t maxima ́lnı ́poc ̌et variant na jednom c ̌tenı ́, n poc ̌et variant a k parametr algoritmu urc ̌ujı ́cı ́ velikost okna.