Can we use it? On the utility ofde novoand reference-based assembly of Nanopore data for plant plastome sequencing

Abstract

The chloroplast genome harbors plenty of valuable information for phylogenetic research.Illumina short-read data is generally used for de novo assembly of whole plastomes. PacBio or Oxford Nanopore long reads are additionally employed in hybrid approaches to enable assembly across the highly similar inverted repeats of a chloroplast genome. Unlike for PacBio, plastome assemblies based solely on Nanopore reads are rarely found, due to their high error rate and non-random error profile. However, the actual quality decline connected to their use has never been quantified. Furthermore, no study has employed reference-based assembly using Nanopore reads, which is common with Illumina data. Using Leucanthemum Mill. as an example, we compared the sequence quality of seven plastome assemblies of the same species, using combinations of two sequencing platforms and three analysis pipelines.In addition, we assessed the factors which might influence Nanopore assembly quality during sequence generation and bioinformatic processing.The consensus sequence derived from de novo assembly of Nanopore data had a sequence identity of 99.59% compared to Illumina short-read de novo assembly. Most of the found errors comprise indels (81.5%), and a large majority of them is part of homopolymer regions.The quality of reference-based assembly is heavily dependent upon the choice of a closeenough reference. Using a reference with 0.83% sequence divergence from the studied species, mapping of Nanopore reads results in a consensus comparable to that from Nanopore de novo assembly, and of only slightly inferior quality compared to a referencebased assembly with Illumina data (0.49% and 0.26% divergence from Illumina de novo). For optimal assembly of Nanopore data, appropriate filtering of contaminants and chimeric sequences, as well as employing moderate read coverage, is essential.Based on these results, we conclude that Nanopore long reads are a suitable alternative to Illumina short reads in plastome phylogenomics. Only few errors remain in the finalized assembly, which can be easily masked in phylogenetic analyses without loss in analytical accuracy. The easily applicable and cost-effective technology might warrant more attention by researchers dealing with plant chloroplast genomes.

Long-read only assembly of Drechmeria coniospora genomes reveals widespread chromosome plasticity and illustrates the limitations of current nanopore methods

“…Here, using only long reads and Canu, we produced the first complete circular mitochondrial genome for D. coniospora and were able to generate chromosome-scale assemblies for the nuclear genome. The rare misassembled contigs, formed by Canu because of single very long chimeric reads, as previously described [16], could be detected by read coverage anomalies and comparisons with unitigs, suggesting that solutions to avoid their creation could be implemented within…”

Yes, we can use it: a formal test on the accuracy of low-pass nanopore long-read sequencing for mitophylogenomics and barcoding research using the Caribbean spiny lobster Panulirus argus

“…Using low-coverage whole genome shot-gun long-read sequencing, most of the pipelines used herein retrieved a complete mitochondrial genome, as shown when these long-read assemblies were compared to a high-coverage 'reference' assembly generated from the same individual but with Illumina short reads. Canu was the only pipeline that failed to assemble and circularize the studied mitochondrial chromosome even when parameters were customized to optimize the assembly of short circular molecules ( [39], see also [40]). Earlier versions of Canu were known to 'have trouble' assembling short circular molecules i.e., plasmids, in datasets with uneven coverage [39].…”

“…Still, the problem persists based on the results from this study. Previous studies using Canu have successfully assembled mitochondrial genomes [32] and larger circular chromosomes (i.e., chloroplasts - [40], bacteria - [22]). Importantly, in this study, Canu not only failed to assemble the mitochondrial genome of P. argus but also took 20-30 times longer to execute compared to the other pipelines that assembled and circularized the studied mitochondrial genome in < 2 h. The current version of Canu automatically detect available resources in a computer or cluster and configures itself to run 'efficiently' using those resources [39].…”

“…Only a slight but detectable decrease in accuracy occurred when the reference genome was from the distantly related species P. versicolor. To compare, a recent study that assembled chloroplasts using nanopore long reads did report a decrease in accuracy when the genome used as a reference belonged to a distantly related species [40]. In this study, obvious differences among assemblies were observed only when extra polishing with the pipeline Medaka was applied, in line with expectations; the pipeline Medaka should result in improved accuracy given the state-of-the-art complex neural network applied by this tool while searching for sequence consensus (https://github.com/nanoporetech/medaka).…”

“…The hierarchical assembly pipeline in Canu comprises four steps: (i) detection of overlaps in the set of sequences, (ii) generation of a corrected sequence consensus, (iii) trimming corrected sequences, and (iv) assembly of trimmed corrected sequences [39]. Canu was executed twice, once with default options (but using genomeSize = 16 k) and a second time with modified parameters (genomeSize = 16 k, correctedError-Rate = 0.134, corOutCoverage = 1000, min-OverlapLength = 300 bp) that was expected to optimize the retrieval of short circular molecule assemblies, including mitochondrial genomes ( [39], see also [40]).…”