Characterization of the De novo Assembly Using Oxford Nanopore Sequencing Data
- 1. Department of Computer Science and Engineering, University of Nebraska-Lincoln, USA
Citation
Liu K (2018) Characterization of the De novo Assembly Using Oxford Nanopore Sequencing Data. Int J Plant Biol Res 6(2): 1087.
EDITORIAL
The MinION of Oxford Nanopore Technology (ONT) is a portable single-molecule sequencer released in 2014. As a portable nanopore-based sequencer [1], with just a USB drive size which can be conveniently used connected to a desktop or laptop using the USB 3.0 or higher interface, MinION contains pores embedded on a membrane that is placed over an electric grid. DNA/RNA molecules sequenced by MinION are basically measured by the ionic flow changes through the pores. As a thirdgeneration commercial sequencing platform product, Nanopore sequencing data can reach up to hundreds of thousands of nucleotides in a single run. Usually an average of 5,000 bp of the product long reads are expected for the DNA sequencing. When the two strands of the target sequence are both successfully basecalled, a consensus will be generated as a more accurate output (the “2D” sequence) compared with forward strand only output (“1D” sequence). Such long read sequencing technology enables great capacity in assembling large and complex genome data compared with using short reads only.
However, such long reads are generally much more errorprone (10~30% error rate) than short-read sequencing technologies such as Illumina, which generally makes it less competent in direct usage such as small structural variations (SNP and InDel) detections and other sequence analysis and applications. Nanopore sequencing shows a pattern of error in base calling being context-specific, for example, small variations of SNP such as TAG->TGG, TAC ->TCG are predominant in ONT. Early reports [2], showed about ~35% error rate in ONT data significantly hindered the wide application of this new technology. Therefore, Nanopore data usually needs to be corrected for the preprocessing. With the development of the base calling software and nanopore chemistry, a significant drop in error rate has been achieved [3]. Both short reads-based and long reads self-based correction strategies can be conducted based on the input data sets. Common error correction tools as nanocorrect [4], nanocorr [5], can be used for the correction of Nanopore data.
The basic data processing pipeline using ONT sequencing data contains: preprocessing, error correction, sequence assembly. For genome assembly, it is reported that using nanopore data only at about 30X genome coverage can be sufficient for assembling some small genomes such as E. coli [4]. Even after error correction, the assembly using Nanopore long reads is still quite challenging now since many existing assemblers were not designed to implement long reads with high error rate. Therefore, some hybrid methods incorporating both long and short reads for genome assembly are developed. The hybrid and non-hybrid de novo assembly strategies are both important depend on the data sets: short read will offer high quality for small region of sequence assembly as well as the scaffolding using paired end information, while Nanopore can recover long repetitive regions which cannot be fully reconstructed using short reads only. The selection between hybrid and non-hybrid assembly methods also depends on other considerations such as the size of the target genome, G+C content bias of the genome composition, the genome complexity (repetitive ratio and multiploidy, etc), also the costs and the bioinformatics analysis to be implemented for the project.
For de novo assembly tools can be used similar to PacBio reads such as PBc R and canu [6]. PBc R is an excellent assembler used in PacBio small and large genome de novo assembly using either hybrid or self-based methods. Canu can also help effectively assemble MinION data into genomes with high sequence accuracy. SPAdes [7] is another assembler used for both simple and multichromosome genomes. Other assemblers using de Bruijn graphbased methods such as Velvet [8] and ABySS [9], as well as using Overlap Layout Consensus (OLC)-based methods such as Celera assembler named CABOG [10]. Also other greedy graph-based package such as SSAKE [11], can also be applied in Nanopore reads de novo assembly. A survey [12] of the benchmarking of those assemblers in their performance using Nanopore long genome reads showed that an ideal strategy of long read assembly should first choose the OLC-based assemblers for a higher initial N50 value & mean and then use the de Bruijn graph-based algorithms for the accuracy improvement. NaS pipeline [13] combines both de Bruijn graphs and OLC approach for the error-free DNA reads assembly.
For plant genome sequence research, breakthroughs using Nanopore technology have proved the competence of long read in genome assembly of generating very large and long contiguous sequences (contigs) of complex genomes. For Oryza coarctata [14], a tetraploid Asian wild rice of approximately 660 Mb in size, a draft genome of multi-chromosome assembled using both hybrid and non-hybrid assembly strategies demonstrated the ability of de novo assembly using Nanopore high-noise single molecule sequencing reads. Arabidopsis thaliana genome is also successfully reconstructed using Nanopore reads in a fast and cost-effective implementation. Other plant pathogen genome assemblies using Nanopore genomic dataset such as Agrobacterium tumefaciens [15], also provided more evidences on long reads being effectively assembled into multi-chromosomal genomes with small number of contigs and high accuracy.
