Assessment of low-coverage nanopore long read sequencing for SNP genotyping in doubled haploid canola (Brassica napus L.)

M M Malmberg,G C Spangenberg,N O I Cogan,H D Daetwyler

doi:10.1038/s41598-019-45131-0

Abstract

Despite the high accuracy of short read sequencing (SRS), there are still issues with attaining accurate single nucleotide polymorphism (SNP) genotypes at low sequencing coverage and in highly duplicated genomes due to misalignment. Long read sequencing (LRS) systems, including the Oxford Nanopore Technologies (ONT) minION, have become popular options for de novo genome assembly and structural variant characterisation. The current high error rate often requires substantial post-sequencing correction and would appear to prevent the adoption of this system for SNP genotyping, but nanopore sequencing errors are largely random. Using low coverage ONT minION sequencing for genotyping of pre-validated SNP loci was examined in 9 canola doubled haploids. The minION genotypes were compared to the Illumina sequences to determine the extent and nature of genotype discrepancies between the two systems. The significant increase in read length improved alignment to the genome and the absence of classical SRS biases results in a more even representation of the genome. Sequencing errors are present, primarily in the form of heterozygous genotypes, which can be removed in completely homozygous backgrounds but requires more advanced bioinformatics in heterozygous genomes. Developments in this technology are promising for routine genotyping in the future.

Highlights

Brassica napus is a recent allotetraploid arising from natural hybridisation events between B. rapa and B. oleracea, with high collinearity to both progenitor genomes[1]
While skim WGR using Long read sequencing (LRS) is probably currently unsuitable for de novo single nucleotide polymorphism (SNP) discovery without correction or substantial sequencing depth, there is the potential that LRS can accurately genotype pre-determined SNP positions, as long as alignment algorithms are able to correctly break and align reads to a reference genome around short insertions and deletions, which are the predominant form of error in Oxford Nanopore Technologies (ONT) minION sequencing[30]
For minION LRS to be realistically used for genotyping purposes, the resulting genotype calls need to contain fewer errors than the overall 10% error rate, this may be tolerable for genomic selection, which has been found to be relatively robust to error rates of up to c. 10%36

Summary

Introduction

Brassica napus is a recent allotetraploid arising from natural hybridisation events between B. rapa and B. oleracea, with high collinearity to both progenitor genomes[1]. Since skim whole genome re-sequencing (skim WGR) using SRS has been found to be highly cost-effective, high-throughput and a relatively accurate genotyping-by-sequencing method[32,33,34,35], LRS must be sufficiently accurate at skim levels without correction in order to compete with current SRS technology. While skim WGR using LRS is probably currently unsuitable for de novo SNP discovery without correction or substantial sequencing depth, there is the potential that LRS can accurately genotype pre-determined SNP positions, as long as alignment algorithms are able to correctly break and align reads to a reference genome around short insertions and deletions, which are the predominant form of error in ONT minION sequencing[30]. This study aimed to i) assess the effect of long reads on misalignment to the reference genome, ii) determine if the overall confidence in the resulting genotypes can be improved by applying various filtering treatments, and iii) describe the nature of discrepancies between Illumina and minION sequencing derived genotypes

Objectives

Methods

Results

Discussion

Conclusion