High quality genome assemblies of Mycoplasma bovis using a taxon-specific Bonito basecaller for MinION and Flongle long-read nanopore sequencing

Nick Vereecke,Jade Bokma,Sebastiaan Theuns,Filip Boyen,Bart Pardon,Freddy Haesebrouck,Hans Nauwynck

doi:10.1186/s12859-020-03856-0

Nick Vereecke, Jade Bokma + Show 5 more

Open Access

https://doi.org/10.1186/s12859-020-03856-0

Copy DOI

Abstract

BackgroundImplementation of Third-Generation Sequencing approaches for Whole Genome Sequencing (WGS) all-in-one diagnostics in human and veterinary medicine, requires the rapid and accurate generation of consensus genomes. Over the last years, Oxford Nanopore Technologies (ONT) released various new devices (e.g. the Flongle R9.4.1 flow cell) and bioinformatics tools (e.g. the in 2019-released Bonito basecaller), allowing cheap and user-friendly cost-efficient introduction in various NGS workflows. While single read, overall consensus accuracies, and completeness of genome sequences has been improved dramatically, further improvements are required when working with non-frequently sequenced organisms like Mycoplasma bovis. As an important primary respiratory pathogen in cattle, rapid M. bovis diagnostics is crucial to allow timely and targeted disease control and prevention. Current complete diagnostics (including identification, strain typing, and antimicrobial resistance (AMR) detection) require combined culture-based and molecular approaches, of which the first can take 1–2 weeks. At present, cheap and quick long read all-in-one WGS approaches can only be implemented if increased accuracies and genome completeness can be obtained.ResultsHere, a taxon-specific custom-trained Bonito v.0.1.3 basecalling model (custom-pg45) was implemented in various WGS assembly bioinformatics pipelines. Using MinION sequencing data, we showed improved consensus accuracies up to Q45.2 and Q46.7 for reference-based and Canu de novo assembled M. bovis genomes, respectively. Furthermore, the custom-pg45 model resulted in mean consensus accuracies of Q45.0 and genome completeness of 94.6% for nine M. bovis field strains. Improvements were also observed for the single-use Flongle sequencer (mean Q36.0 accuracies and 80.3% genome completeness).ConclusionsThese results implicate that taxon-specific basecalling of MinION and single-use Flongle Nanopore long reads are of great value to be implemented in rapid all-in-one WGS tools as evidenced for Mycoplasma bovis as an example.

Highlights

Implementation of Third-Generation Sequencing approaches for Whole Genome Sequencing (WGS) all-in-one diagnostics in human and veterinary medicine, requires the rapid and accurate generation of consensus genomes
These results implicate that taxon-specific basecalling of MinION and single-use Flongle Nanopore long reads are of great value to be implemented in rapid all-in-one WGS tools as evidenced for Mycoplasma bovis as an example
Improved reference‐based and de novo assemblies using an M. bovis trained Bonito basecalling algorithm Performing basecalling with a custom-trained Bonito v.0.1.3 model, using taxon-specific training data resulted in a significantly increased final consensus accuracy for reference-based genomes (Q45.2) (Fig. 1a)

Summary

Introduction

Implementation of Third-Generation Sequencing approaches for Whole Genome Sequencing (WGS) all-in-one diagnostics in human and veterinary medicine, requires the rapid and accurate generation of consensus genomes. Different single-molecule long-read sequencing approaches have become available nowadays, requiring a relatively small upfront financial investment as compared to Illumina short-read sequencing instrument costs [4] The latter is still the predominant Generation Sequencing (NGS) technology, as highly accurate reads are generated (0.1% substitutions for MiSeq) in comparison to the long-read Oxford Nanopore Technologies (ONT) approach (5% for R9.4.1 flow cell chemistry; ONT communication March 2020) [4, 5]. Upon the insertion of DNA in the nanoscale pore using a secondary motor protein, blockage of the pore-channel allows to measure differential voltages for each k-mer placed within the pore This results in raw current signals, known as squiggle spaces, which allow interpretation of the k-mer sequence stretch blocking the nanopore at a specific moment. This is of great value for various approaches that require squiggle (re-)alignment including basecalling, adaptive sequencing, methylation calling, and their respective training workflows [8, 9]

Objectives

Methods

Results

Discussion

Conclusion