The impact of applying various de novo assembly and correction tools on the identification of genome characterization, drug resistance, and virulence factors of clinical isolates using ONT sequencing
BMC Biotechnology volume 23, Article number: 26 (2023)
Oxford Nanopore sequencing technology (ONT) is currently widely used due to its affordability, simplicity, and reliability. Despite the advantage ONT has over next-generation sequencing in detecting resistance genes in mobile genetic elements, its relatively high error rate (10–15%) is still a deterrent. Several bioinformatic tools are freely available for raw data processing and obtaining complete and more accurate genome assemblies. In this study, we evaluated the impact of using mix-and-matched read assembly (Flye, Canu, Wtdbg2, and NECAT) and read correction (Medaka, NextPolish, and Racon) tools in generating complete and accurate genome assemblies, and downstream genomic analysis of nine clinical Escherichia coli isolates. Flye and Canu assemblers were the most robust in genome assembly, and Medaka and Racon correction tools significantly improved assembly parameters. Flye functioned well in pan-genome analysis, while Medaka increased the number of core genes detected. Flye, Canu, and NECAT assembler functioned well in detecting antimicrobial resistance genes (AMR), while Wtdbg2 required correction tools for better detection. Flye was the best assembler for detecting and locating both virulence and AMR genes (i.e., chromosomal vs. plasmid). This study provides insight into the performance of several read assembly and read correction tools for analyzing ONT sequencing reads for clinical isolates.
The rapid development of whole-genome sequencing tools expanded their usage to sequence the whole genome of species from small single cells to large and complex species . Besides, the reduction in sequencing costs and simplicity in library preparations allowed a worldwide distribution of these sequencing tools, where low- and mid-income countries have great access to such advanced tools [2, 3]. Next-generation sequencing (NGS) technology has wholly revolutionized genome analysis. NGS is relatively timesaving in that it can promptly verify a sample sequence type and the presence of critical genes [4, 5]. The production of millions of short reads (mostly < 150 bp in length) during NGS and a low error rate (< 2%) makes this technology highly accurate and reliable in identifying single nucleotide polymorphisms (SNPs) and understanding population structures . Despite the advantages of using short reads, especially with sequencing accuracy, the major shortcomings of using short reads are the failure to assemble complete genomic structures and poor gene organization even with special library preparations that intend to increase the depth of sequencing coverage, and the inability to resolve all genomic repeats [7, 8]. The location of resistance genes is critical epidemiologically -especially for public health laboratories- and unresolved/incomplete assembled genomes cannot indicate whether the resistant gene is located on the chromosome or the mobile genetic elements (MGEs) [9, 10]. Another drawback of NGS is the possible introduction of a biased nucleotide variance during the PCR amplification step of library preparation .
To conquer the hindrance of unresolved assemblies and inaccurate repetitive region sequencing, Oxford Nanopore Sequencing Technology (ONT), a third-generation sequencing technology, generates long reads that exceed the length of repeated regions resulting in complete genome assemblies with more accurate gene locations (i.e., chromosomal vs. MGEs) [9, 12]. ONT generates reads ranging from 500 bp to 2.3 Mb, with 10–30 kb genomic libraries being the most prevalent . The main limitation of ONT is the relatively higher error rate (~ 10–15%) compared to NGS, though the technology is constantly improving . Computational efforts were made to develop assembly and correction tools to transform raw ONT signals into completely assembled genomes, taking into consideration both the high error rates and the length of reads. The assembly algorithms and correction tools aim to reduce the high error rates.
Several long-read assemblers are freely available online. These assemblers use different algorithms to generate the best consensus sequences in combination with a secondary correction tool. Flye is a long-read de novo assembler based on a generalized Bruijn Graph. Flye combines multiple disjoint genomic segments, connects all error-prone disjointigs into a single string, and outputs an accurate assembly graph . Flye package can be called a ‘complete pipeline’ as this tool converts raw ONT reads into corrected consensus sequences . Canu is an assembler based on the over-layout-consensus algorithm (OLC) designed for reconstructing noisy long reads . Canu operates in three phases: correction, trimming, and assembly. Canu improves base calling error in the correction phase by trimming low-quality bases and adapters in the trimming phase, and finally arranges the reads into contigs to generate both consensus sequences and graphs of alternate paths in the assembly phase [17, 18]. Wtdbg2 is another assembler that is based on the OLC algorithm. Wtdbg2 cleaves long reads into 1024 segments, merges similar segments into a vertex, and connects vertices based on segment adjacency on reads producing a fuzzy de Bruijn graph . A major advantage of using Wtdbg2 is its speed of assembly, which can be 10–17 times faster than other assemblers such as Canu to produce comparable consensus sequences. However, Wtdbg2 assembles raw reads without an error correction step . NECAT is a novel two-stage assembler for noisy long-reads designed to overcome complex errors in Nanopore reads. NECAT corrects and assembles raw reads into high-quality consensus sequences relatively faster than Canu . To generate genome assemblies with a reduced error rate, read correction tools -or polishing tools- are often required. Though the correction step takes longer than genome assembly, the ‘corrected/polished’ assembly is more complete and is much improved . For example, NextPolish contains two core modules and uses a stepwise mode to fix base errors (SNV/Indel) , while Racon is intended as a standalone consensus module to correct contigs generated by assembly methods .
The evaluation and benchmarking of various long-read assembly tools using reference laboratory strains have been widely explored in the literature [11, 18, 24]. However, benchmarking the usage of assembly and read correction tools in clinical isolates has not been sufficiently investigated. In this study, we assessed and compared the capability of mix-and-matched assembly and read correction tools in generating complete and accurate genome assemblies, and downstream genomic analysis regarding strain and serotype identification, annotation, detection of antimicrobial resistance genes, plasmid finding, and virulence potential of nine clinical Escherichia coli isolates.
Materials and methods
The sequencing reads were submitted to EMBL's European Bioinformatics Institute and are available online with accession numbers ERR10468513- ERR10468521 (Suppl. S1 Table S1) available at: https://www.ebi.ac.uk/ena/browser/view/PRJEB57325.
Bacterial isolates, library preparation, MinION sequencing, and reads preparation
E. coli isolates (n = 9) were obtained from rectal swabs of pregnant women admitted to the Gynecology ward in Farwaniya Hospital, Kuwait. All methods and ethical approvals were obtained and performed in accordance with the Ethical Committees of the Health Sciences Centre, Kuwait University, and the Ministry of Health, Kuwait. The patients/participants (or their legal guardians) provided their written informed consent to participate in this study. These bacterial isolates were grown overnight at 37 °C on selective agar, and single colonies were suspended in phosphate-buffered saline (PBS). Bacterial genomic DNA was purified using Monarch genomic DNA purification kit (New England BioLabs) as follows. Bacterial cell pellets from PBS were generated by centrifugation at 10,000 g for 5 min. The pellet was resuspended with 1 µl of proteinase K and 3 µl of RNAse, and then 100 µl of cell lysis buffer was added. The samples were vortexed and incubated for 5 min at 56 °C with agitation. Genomic DNA was extracted from the lysed samples following Part 2 of the kit protocol for cultured cells from the binding step onwards. The extracted DNA was checked for quality and quantity using a spectrophotometer (Nanodrop, ThermoFisher Scientific), and a fluorometer (Qubit, ThermoFisher Scientific), respectively.
Oxford Nanopore Ligation Sequencing kit (SQK-LSK109) with Native Barcoding Expansion 1–12 (EXP-NBD104) was used for library preparation. Following the kits protocols, 1 µg of the isolated genomic DNA was treated with end-repair/dA tailing module, and then DNA was eluted after AMPure XP bead clean up. The genomic DNA was barcoded following the Native Barcoding Expansion 1–12 (EXP-NBD104) protocol and then cleaned with AMPure XP beads. The eluted barcoded genomic DNA was pooled to 65 µl and used for adapter ligation. The final library contained 29.8 ng DNA and 50 fmol of the library was loaded onto an R9.4 flow cell. The sequencing was performed using the MiniON Mk1C device with a flow cell (FLO-MI106D) – containing 875 available pores) following the user manual (Suppl. S1 Table S2). The run proceeded for the full 48 h.
The built-in Guppy v4.0.11 (https://community.nanoporetech.com) base-called and demultiplexed the fast5 reads and output fastq files. Barcodes and adapter sequences were then trimmed from reads using Porechop (v0.2.1, https://github.com/rrwick/Porechop). The resulting 9 demultiplexed, barcode-free read sets were deposited under accession numbers ERR10468513- ERR10468521 (Suppl. S1 Table S1). NanoPlot and Nanofilt were used to assess the reads quality and filter reads quality to gather reads with q score > 8 (Suppl. S1 Table S3) .
De novo assembly, read correction, and assembly assessment
Four de novo genome assemblers: Flye (version 2.8.3-b1695), Canu (version 1.9), Wtdbg2 (version 3.0), and NECAT (version 0.0.1), and three read correction tools: Medaka (version 0.11.0), NextPolish (version 1.4.1), and Racon (version 1.4.10) were selected for assembling and correcting long reads generated by ONT [15, 17, 19, 20, 22, 23, 26]. The reads for each isolate were assembled using each of above mentioned de novo assemblers with default settings. After each assembly, the consensus files generated followed one round of read correction using Medaka, NextPolish or Racon with default settings. The running time and CPU usage are available in Suppl. S1 Table S4. This resulted in generating 12 corrected assemblies per read set. The consensus sequences generated from each corrected assembly underwent a similarity sequence against the nucleotide database NCBI Basic Local Alignment Search Tool using BLAST + Command Line Application tool v.2.12.0 for large contigs (> 1,000,000 bp) . The E. coli serotype determination was performed using ECTyper (version 1.0.0) . The quality of each corrected assembly was assessed using QUAST (version 5.0.2)  –using the LG parameter- comparing them with appropriate reference genome for each isolate (based on BLAST + results). The total length (bp), number of contigs, GC% and total aligned sequence (bp) were evaluated in each corrected assembly.
Identification of genes annotation, antimicrobial resistance genes, plasmids, and virulence genes
The E. coli genomes were annotated using Prokka (version 1.14.5)  and the generated GFF files were used as input for pangenome inference using Roary (version 3.13.0)  to generate the core- (genes present in all analyzed isolates), shell- (genes present in the majority of genomes), and cloud- (genes present in the minority of the genomes) genes. Antimicrobial resistance genes were detected using staramr (version 0.7.2)  against known gene sequences in the ResFinder database  with 98% minimum identity and 60% minimum coverage and using Resistance Gene Identifier (RGI) strict criteria . Plasmids were identified using staramr against known plasmid sequences in the PlasmidFinder database  with 98% minimum identity and 60% minimum coverage. Virulence genes were identified using ABRicate (version 2.0) integrated with Virulence Factors Database (VFDB) with 98% minimum identity and 60% minimum coverage [36, 37], and Venn diagrams were constructed using online tool https://bioinformatics.psb.ugent.be/webtools/Venn/.
Wilcoxon signed-rank test was performed using GraphPad Prism (California, USA) (version 9.4.1) to determine whether significant differences (p < 0.05) existed between assembled genomes with and without using read correction tools in total length (bp), GC%, total aligned (bp), and indels.
Results and discussion
Oxford Nanopore Sequencing technology generates long-reads that overcome NGS limitations, especially when sequencing repeated tandems. However, the performance and optimization of read assembly and read correction tools of ONT long reads still warrant further investigation. In this study, we aimed to evaluate four read assembly (Flye, Canu, Wtdbg2 and NECAT) and three read correction (Medaka, NextPolish and Racon) tools in assembling nine clinical E. coli isolates. Since a reference strain was not available for this study, we evaluated assembly accuracy by aligning assemblies to a reference genome obtained from the NCBI database (NC_002655.2, NZ_CP017979.1, NZ_CP051263.1, and NC_007779.1), and therefore, the primary purpose was to assess read assembly and correction tools in gene structure and completeness. To evaluate the effect of different de novo assembly and correction tools on the strain and serotype identification, we used both BLAST + and ECTyper tools. The data presented in Tables 1 and 2 indicate that all assembled genomes belongs to the same strains and serotypes (MLST prediction is presented in Suppl. S1 Table S5). The identified strains from BLAST + were then used as references in QUAST for evaluating assembly and correction tools (Table 3). The four de novo assembly tools could generate consensus files for all nine isolates except for NECAT, which failed to assemble the genomes of two samples (barcodes 02 and 08) due to the high number of short reads and/or the high number of contigs. The assembled genomes’ total length (bp) was higher when using Flye and Canu (Table 3). The larger genome assemblies produced by read assembly tools show the ONT advantage of sequencing organisms with moderate GC content. In general, all corrected assemblies improved total genome length and were significantly (p < 0.05) improved when using Medaka and Racon as read correction tools but not NextPolish (Fig. 1). Similarly, Wang et al. have reported significant improvement in genome sizes after using read correction tools . In their study, the reads were corrected by up to 57%. In this study, Wtdbg2 generated the shortest assembled genomes, followed by NECAT (Table 3). The performance of Wtdbg2 and NECAT for assembling ONT reads has been controversial. While several studies have suggested that Wtdbg2 and NECAT perform well in assembling good- and low-quality ONT reads, especially after correcting with Medaka [12, 18, 24], the same was not noticed in this study. The reason could be the methods of DNA extraction used, library preparation and/or the bacteria being sequenced. The number of contigs was not significantly affected by read correction tools and were all the same before and after read corrections (except for barcode 12). Most read correction tools performed well in terms of GC content compared to non-corrected assemblies. Only in barcode 12, the number of GC% was significantly (p < 0.05) lower in non-corrected reads. However, Medaka had significantly (p < 0.05) lower GC%. The number of indels was significantly (p < 0.05) lower when using Flye (with and without using correction tools) and significantly (p < 0.05) higher when using Wtdbg2 (Table 3, Fig. 1). The use of Medaka and Racon read correction tools significantly (p < 0.05) lowered the number of indels when using Wtdbg2. The usage of read correction tools to lower indels number was also detected by others . The relatively high number of indels is continuously noticed with ONT sequencing, which could introduce errors (such as a stop codon) that affects gene annotation [7, 39].
A total of 15 assembled genomes (12 corrected and three non-corrected) per sample were included in the pan-genome analysis, which displayed core and accessory (shell and cloud) genes. Flye was the most effective assembler for the pan-genome analysis, followed by NECAT and Canu (Fig. 2a and b). Non-corrected Flye assemblies consisted of total genes of 33,257, 1,560 (4.6%) of which were core genes and 31,697 (95%) were accessory genes. The total number of indels was the lowest in genomes assembled by Flye. Since the total number of indels detected by read assembly and correction tools affects gene annotation in which the high read errors produce more misannotated genes and the number of accessory genes, this could explain the good performance of Flye in detecting the highest core genes . Interestingly, genomes assembled by Canu, Wtdbg2, and NECAT had more core genes when corrected by Medaka compared to non-corrected assemblies (Fig. 2a). Although Wtdbg2 showed the highest number of total genes 48,132, it showed the least number of core genes (16 – 0.03%). This could be due to the inaccurate genome size produced by Wtdbg2 and a high number of indels detected. Genomes assembled by NECAT (with and without correction) had the lowest number of total and accessory genes (Fig. 2b). The use of Racon as a read correction tool for genomes assembled by Flye and Canu increased the number of total and accessory genes (Fig. 2b). However, this was not noticed in genomes assembled by Wtdbg2 and NECAT.
A major advantage of ONT sequencing is the rapid identification of antimicrobial resistance genes, plasmids, and virulence genes in bacterial genomes [40, 41]. Besides, the long-reads generated allow the detection of the presence/absence of antimicrobial and virulence genes and their architectures i.e., chromosomal vs. plasmid . In this study, we investigated nine clinical E. coli isolates. Antimicrobial resistance genes were detected using two independent tools: staramr (ResFinder) and RGI, the plasmid detection by PlasmidFinder, and virulence genes by Abricate. The results obtained from these tools did not follow a particular pattern, however, some read and correction tools performed better than others. Flye, Canu and NECAT read assembly tools performed well in detecting antimicrobial resistance genes when using staramr with Flye being the best assembler to identify resistant genes in genomes regardless of the correction tool used followed by Canu (Fig. 3). Both Flye and Canu were able to identify resistance genes by 98–99% or 100% identity while Wtdbg2 and NECAT missed these genes (barcodes 02, 04, 08, 09, and 12). Interestingly, only Wtdbg2/Racon was able to detect resistance in chloramphenicol in barcode 09, and ampicillin (AMP), erythromycin (ERY), lincomycin (L) and streptomycin (STR) in barcode 11. The RGI analysis followed staramr results to a certain degree. For example, Wtdbg2 did not detect the presence of baeR, baeS, and CMY-136 genes -antibiotic efflux genes that confer resistance to multiple drug classes- in barcode 01, and most resistance genes in barcode 02 (Fig. 4). However, this was improved when using Medaka and Racon read correct tools (Fig. 4). Although hybrid assembly between short- and long-reads is recommended for more accurate antimicrobial resistance profiling , we have noticed that usage of Medaka and Racon read correction tools enhanced Wtdgb2 prediction in which 133 genes conferring resistance to different antibiotic classes were detected across all samples (Fig. 4).
The plasmid detection results using PlasmidFinder were also inconsistent. Generally, Canu performed the best predicting most of the plasmids in all samples, however, it failed to predict the presence of IncX4 plasmid in barcode 09 and IncHI2 plasmid in barcode 11, which was only detected by Flye and Wtdgb2 after correction with Medaka and Racon but not NextPolish. Surprisingly, Wtdbg2 (with and without correction) could not detect the presence of all Col plasmids (ColBS512, ColMP18, and Col156) as well as IncFI1 and IncFII, while Flye failed to detect ColBS156, IncFI1, and IncHI1. In this study, we detected antimicrobial resistance genes detection on plasmids in four isolates (Table 4). For example, In barcode 01 several resistance genes blaCMY-7, aph(6), blaTEM-34, dfrA14, mph(A), strA, and sul2 were detected on two plasmids, IncI1 And IncQ1. Not all read assembly and correction tools were able to detect the presence of these genes on plasmids. Reads assembled by Flye and corrected by NextPolish could not detect blaTEM, dfreA, and mph. In barcode 03, reads assembled by Wtdbg2 could not detect the sul2 gene when using staramr (ResFinder), however, the gene was detected when using RGI (Figs. 4 and 5). Contradictory, the same gene was detected in barcode 09 by reads assembled by Canu and corrected by NextPolish when using staramr (ResFinder) but not RGI. Although long reads have shown better detection of plasmids than short reads , the inconsistency of plasmid detection when using multiple read and correction tools was also noted in other studies [11, 38]. George et al. reported that regardless of the read assembly tool used for long reads, some small-sized plasmids were missed and only retained when using hybrid assembly . This means that library preparations, bioinformatic tools, and/or sequencing technology for long-reads still need to be improved for accurate plasmid detection.
Identifying virulence factors is critical in understanding E. coli pathogenicity that may impact human health . In this study, we did not notice a significant difference in virulence factors detected after using different read correction tools when using ABRicate (Suppl. S2). Among the nine clinical E. coli isolates differences in virulence detection were noticed in four samples (barcodes 02, 04, 08, and 09) which followed a particular pattern (Fig. 6). Flye performed as the best read assembly tool, followed by Canu, Wtdbg2, and then NECAT. Flye was also able to detect virulence genes on the plasmids. In barcode 08 the gene encoding for enterotoxin senB was detected on Col156 plasmid. Besides, Flye, Canu, and NECAT were able to detect the iroB, iroC, iroD, iroE, iroN, genes on the IncFIA plasmid in barcode 09. Although a reference strain was included in this analysis, the clinical strains may not necessarily match the total number of virulence factors detected in the reference strain. A shortcoming of this study is the unavailability of another sequencing method as a reference. Therefore, we could not be definitive regarding the number of virulence factors nor if any gene was lost during library preparation. The usage of long reads is arguably better for the rapid detection of virulence factors. Obtaining a circular/closed genome with fewer read errors is much more robust in outbreak surveillance and investigations [40, 45, 46]. Unfortunately, although improved, different virulence factors were detected when applying multiple bioinformatic tools .
Third-generation sequencing tools, such as ONT, are acceptable options for whole-genome sequencing especially in low to mid income countries due to their affordability and simplicity. The relatively higher per-read error rate of ONT, which necessitates different assembly and correction approaches to transform raw signals into completely assembled genomes can be reduced by using freely available read assembly and read correction tools. There is necessity of benchmarking real data sets from clinical isolates. In this study, we found that the use of mix-and-matched read assembly and read correction tools can lead to significant differences in total bacterial length, AMR detection, and plasmid and virulence factor identification.
Mobile genetic element
Next generation sequencing
Oxford Nanopore Technology
Resistance gene identifier
Single nucleotide polymorphism
Lander ES, Linton LM, Birren B, Nusbaum C, Zody MC, Baldwin J, Devon K, Dewar K, et al. Initial sequencing and analysis of the human genome. Nature. 2001;409:860–921. https://doi.org/10.1038/35057062.
Loman NJ, Pallen MJ. Twenty years of bacterial genome sequencing. Nat Rev Microbiol. 2015;13:787–94. https://doi.org/10.1038/nrmicro3565.
Kingsmore SF, Lantos JD, Dinwiddie DL, Miller NA, Soden SE, Farrow EG, Saunders CJ. Next-generation community genetics for low- and middle-income countries. Genome Med. 2012;4:25. https://doi.org/10.1186/gm324.
McCombie WR, McPherson JD, Mardis ER. Next-generation Sequencing Technologies. Cold Spring Harb Perspect Med. 2018;9. https://doi.org/10.1101/cshperspect.a036798.
Besser J, Carleton HA, Gerner-Smidt P, Lindsey RL, Trees E. Next-generation sequencing technologies and their application to the study and control of bacterial infections. Clin Microbiol Infect. 2018;24:335–41. https://doi.org/10.1016/j.cmi.2017.10.013.
Heydari M, Miclotte G, Demeester P, Van de Peer Y, Fostier J. Evaluation of the impact of illumina error correction tools on de Novo Genome Assembly. BMC Bioinformatics. 2017;18(1):374. https://doi.org/10.1186/s12859-017-1784-8.
George S, Pankhurst L, Hubbard A, Votintseva A, Stoesser N, Sheppard AE, Mathers A, et al. Resolving plasmid structures in Enterobacteriaceae using the minion nanopore sequencer: Assessment of minion and minion/illumina hybrid data assembly approaches. Microb Genom. 2017;3(8):e000118. https://doi.org/10.1099/mgen.0.000118.
Salzberg SL, Phillippy AM, Zimin A, Puiu D, Magoc T, Koren S, Treangen TJ, et al. Gage: a critical evaluation of genome assemblies and assembly algorithms. Genome Res. 2011;22:557–67. https://doi.org/10.1101/gr.131383.111.
Ashton PM, Nair S, Dallman T, Rubino S, Rabsch W, Mwaigwisya S, Wain J, et al. Minion Nanopore sequencing identifies the position and structure of a bacterial antibiotic resistance island. Nat Biotechnol. 2014;33:296–300. https://doi.org/10.1038/nbt.3103.
Guo S, Aung KT, Tay MYF, Seow KL, Ng LC, Schlundt J. Extended-spectrum β-lactamase-producing Proteus Mirabilis with multidrug resistance isolated from Raw Chicken in Singapore: Genotypic and phenotypic analysis. J Glob Antimicrob Resist. 2019;19:252–4. https://doi.org/10.1016/j.jgar.2019.10.013.
Juraschek K, Borowiak M, Tausch SH, Malorny B, Käsbohrer A, Otani S, Schwarz S, et al. Outcome of different sequencing and assembly approaches on the detection of plasmids and localization of antimicrobial resistance genes in commensal Escherichia coli. Microorganisms. 2021;9:598. https://doi.org/10.3390/microorganisms9030598.
Wang Y, Zhao Y, Bollas A, Wang Y, Au KF. Nanopore sequencing technology, Bioinformatics and Applications. Nature Biotechnol. 2021;39:1348–65. https://doi.org/10.1038/s41587-021-01108-x.
Amarasinghe SL, Su S, Dong X, Zappia L, Ritchie ME, Gouil Q. Opportunities and challenges in long-read sequencing data analysis. Genome Biol. 2020;21(1):30. https://doi.org/10.1186/s13059-020-1935-5.
Delahaye C, Nicolas J. Sequencing DNA with nanopores: Troubles and biases. PLOS ONE. 2021;16(10):e0257521. https://doi.org/10.1371/journal.pone.0257521.
Kolmogorov M, Yuan J, Lin Y, Pevzner PA. Assembly of long, error-prone reads using repeat graphs. Nature Biotechnol. 2019;37:540–6. https://doi.org/10.1038/s41587-019-0072-8.
Fenderglass/Flye: De novo assembler for single molecule sequencing reads using repeat graphs https://github.com/fenderglass/Flye (accessed June, 2022).
Koren S, Walenz BP, Berlin K, Miller JR, Bergman NH, Phillippy AM. Canu: Scalable and accurate long-read assembly via adaptive k-mer weighting and repeat separation. Genome Res. 2017;27:722–36. https://doi.org/10.1101/gr.215087.116.
Cherukuri Y, Janga SC. Benchmarking of de novo assembly algorithms for Nanopore Data reveals optimal performance of OLC approaches. BMC Genomics. 2016;17:95–105. https://doi.org/10.1186/s12864-016-2895-8.
Ruan J, Li H. Fast and accurate long-read assembly with WTDBG2. Nat Methods. 2019;17:155–8. https://doi.org/10.1038/s41592-019-0669-3.
Chen Y, Nie F, Xie S-Q, Zheng Y-F, Dai Q, Bray T, Wang Y-X, et al. Efficient assembly of nanopore reads via highly accurate and intact error correction. Nat Commun. 2021;12(1):60.
Wang J, Chen K, Ren Q, Zhang Y, Liu J, Wang G, Liu A, et al. Systematic comparison of the performances of de Novo genome assemblers for oxford nanopore technology reads from piroplasm. Front Cell Infect Microbiol. 2021;11:696669. https://doi.org/10.3389/fcimb.2021.696669.
Hu J, Fan J, Sun Z, Liu S. NextPolish: a fast and efficient genome polishing tool for long-read assembly. Bioinformatics. 2019;36:2253–5. https://doi.org/10.1093/bioinformatics/btz891.
Vaser R, Sović I, Nagarajan N, Šikić M. Fast and accurate de novo genome assembly from long uncorrected reads. Genome Res. 2017;27:737–46. https://doi.org/10.1101/gr.214270.116.
Wick RR, Holt KE. Benchmarking of long-read assemblers for prokaryote whole genome sequencing. F1000Res. 2021;8:2138.
De Coster W, D’Hert S, Schultz DT, Cruts M, Van Broeckhoven C. NanoPack: Visualizing and processing long-read sequencing data. Bioinformatics. 2018;34:2666–9. https://doi.org/10.1093/bioinformatics/bty149.
Nanoporetech Nanoporetech/medaka: Sequence correction provided by ONT Research https://github.com/nanoporetech/medaka (accessed June 9, 2022).
Camacho C, Coulouris G, Avagyan V, Ma N, Papadopoulos J, Bealer K, Madden TL. BLAST+: Architecture and applications. BMC Bioinformatics. 2009;10:421. https://doi.org/10.1186/1471-2105-10-421.
Bessonov K, Laing C, Robertson J, Yong I, Ziebell K, Gannon VP, Nichani A, et al. ECTyper: In silico Escherichia coli serotype and species prediction from raw and assembled whole-genome sequence data. Microb Genom. 2021;7(12):000728. https://doi.org/10.1099/mgen.0.000728.
Mikheenko A, Prjibelski A, Saveliev V, Antipov D, Gurevich A. Versatile genome assembly evaluation with Quast-LG. Bioinformatics. 2018;34:i142–50. https://doi.org/10.1093/bioinformatics/bty266.
Seemann T. Prokka: rapid prokaryotic genome annotation. Bioinformatics. 2014;30:2068–9. https://doi.org/10.1093/bioinformatics/btu153.
Page AJ, Cummins CA, Hunt M, Wong VK, Reuter S, Holden MTG, Fookes M, Falush D, Keane JA, Parkhill J. Roary: rapid large-scale prokaryote pan genome analysis. Bioinformatics. 2015;31:3691–3. https://doi.org/10.1093/bioinformatics/btv421.
Bharat A, Petkau A, Avery BP, Chen JC, Folster JP, Carson CA, Kearney A, Nadon C, Mabon P, Thiessen J, Alexander DC, Allen V, El Bailey S, Bekal S, German GJ, Haldane D, Hoang L, Chui L, Minion J, Zahariadis G, Domselaar GV, Reid-Smith RJ, Mulvey MR. Correlation between phenotypic and in silico detection of antimicrobial resistance in Salmonella enterica in Canada using staramr. Microorganisms. 2022;10:292. https://doi.org/10.3390/microorganisms10020292.
Zankari E, Hasman H, Cosentino S, Vestergaard M, Rasmussen S, Lund O, Aarestrup FM, et al. Identification of acquired antimicrobial resistance genes. J Antimicrob Chemother. 2012;67:2640–4. https://doi.org/10.1093/jac/dks261.
Alcock BP, Raphenya AR, Lau TT, Tsang KK, Bouchard M, Edalatmand A, Huynh W, et al. Card 2020: Antibiotic resistome surveillance with the comprehensive antibiotic resistance database. Nucleic Acids Res. 2019;48(D1):D517–25. https://doi.org/10.1093/nar/gkz935.
Carattoli A, Zankari E, García-Fernández A, Voldby Larsen M, Lund O, Villa L, et al. in silico detection and typing of plasmids using PlasmidFinder and plasmid multilocus sequence typing. Antimicrob Agents Chemother. 2014;58:3895–903. https://doi.org/10.1128/AAC.02412-14.
Tseemann Tseemann/abricate: Mass screening of contigs for antimicrobial and virulence genes https://github.com/tseemann/abricate (accessed July 30, 2022).
Chen L, Zheng D, Liu B, Yang J, Jin QVFDB. Hierarchical and refined dataset for big data analysis—10 years on. Nucleic Acids Res. 2016;2015:44.
Goldstein S, Beka L, Graf J, Klassen JL. Evaluation of strategies for the assembly of diverse bacterial genomes using minion long-read sequencing. BMC Genomics. 2019;20(1):23.
Sović I, Križanović K, Skala K, Šikić M. Evaluation of hybrid and non-hybrid methods forde novo assembly of Nanopore reads. Bioinformatics. 2016;32:2582–9. https://doi.org/10.1093/bioinformatics/btw237.
González-Escalona N, Allard MA, Brown EW, Sharma S, Hoffmann M. Nanopore sequencing for fast determination of plasmids, phages, virulence markers, and antimicrobial resistance genes in shiga toxin-producing Escherichia coli. PLOS ONE. 2019;14(7):e0220494. https://doi.org/10.1371/journal.pone.0220494.
Jønsson R, Struve C, Boll EJ, Boisen N, Joensen KG, Sørensen CA, Jensen BH, Scheutz F, Jenssen H, Krogfelt KA. A novel paa virulence plasmid encoding toxins and two distinct variants of the fimbriae of enteroaggregative Escherichia coli. Front Microbiol. 2017;8:263. https://doi.org/10.3389/fmicb.2017.00263.
Greig DR, Dallman TJ, Hopkins KL, Jenkins C. Minion Nanopore sequencing identifies the position and structure of bacterial antibiotic resistance determinants in a multidrug-resistant strain of enteroaggregative Escherichia coli. Microb Genom. 2018;4(10):e000213. https://doi.org/10.1099/mgen.0.000213.
Su M, Satola SW, Read TD. Genome-based prediction of bacterial antibiotic resistance. J Clin Microbiol. 2019;57(3):e01405–18. https://doi.org/10.1128/JCM.01405-18.
Khezri A, Avershina E, Ahmad R. Hybrid Assembly provides improved resolution of plasmids, antimicrobial resistance genes, and virulence factors in Escherichia coli and Klebsiella pneumoniae clinical isolates. Microorganisms. 2021;9(12):2560. https://doi.org/10.3390/microorganisms9122560.
Turton JF, Payne Z, Coward A, Hopkins KL, Turton JA, Doumith M, Woodford N. Virulence genes in isolates of Klebsiella pneumoniae from the UK during 2016, including among carbapenemase gene-positive hypervirulent K1-ST23 and ‘non-hypervirulent’ types ST147, ST15 and ST383. J Med Microbiol. 2018;67:118–28. https://doi.org/10.1099/jmm.0.000653.
Ruan Z, Wu J, Chen H, Draz MS, Xu J, He F. hybrid genome assembly and annotation of a pandrug-resistant klebsiella pneumoniae strain using nanopore and illumina sequencing. Infect Drug Resist. 2020;13:199–206. https://doi.org/10.2147/IDR.S240404.
Chen Z, Erickson DL, Meng J. Benchmarking Hybrid Assembly approaches for genomic analyses of bacterial pathogens using Illumina and Oxford Nanopore sequencing. BMC Genomics. 2020;21(1):1–21. https://doi.org/10.1186/s12864-020-07041-8.
This research was funded internally by the Microbiology Department, Faculty of Medicine, Kuwait university. The funding was independent of the study design and delivery.
Ethics approval and consent to participate
All methods and ethical approvals were obtained and performed in accordance with the Ethical Committees of the Health Sciences Centre, Kuwait University, and the Ministry of Health, Kuwait. The patients/participants (or their legal guardians) provided their written informed consent to participate in this study.
Consent for publication
The authors declare no conflict of interest.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Table S1. Accession numbers of read sets. Table S2. Overview of the sequencing run. Table S3. Nanoplot statistics before and after filtering. Table S4. Benchmark the running time and CPU usage of read trimming, assembly, read correction, quality control and tertiary analysis tools. Table S5. MLST prediction by staramr.
About this article
Cite this article
Safar, H.A., Alatar, F., Nasser, K. et al. The impact of applying various de novo assembly and correction tools on the identification of genome characterization, drug resistance, and virulence factors of clinical isolates using ONT sequencing. BMC Biotechnol 23, 26 (2023). https://doi.org/10.1186/s12896-023-00797-3