DOI: 10.1128/CMR.00016-17
Article Figures & Data
Figures
- FIG 1
Simplified SCS construction.
- FIG 2
Simplified k-mer construction during de Bruijn graph assembly.
- FIG 3
WGS outbreak analysis tools. Different steps in the analysis of WGS data are shown in orange (assembly, genomic characterization, comparative genomics, and phylogeny). Analysis tools are grouped by the analysis step that they perform and are separated by user interface in shades of blue (complete analysis software suites, Web based, and command line).
Tables
- TABLE 1
Performance analysis of sequencing platformsa
Platform Read length (bp) Output (Gb) Coverageb Run time (h) No. of reads Cost per Gb ($) Consumables cost ($) Instrument cost ($) Error rate Dimensions (width × depth × ht) (cm) Source(s) (reference[s]) Sequencing by synthesis Illumina MiniSeq Mid Output 2 × 150c 2.1–2.4c 8.6 17c 14 million–16 millionc 2,584–2,953d 6,201c 55,411c 0.1% in >80% of base callsc 45.6 × 48 × 51.8c Illumina Illumina MiniSeq High Output 1 × 75c 1.7–1.9c 6.8 7c 22 million–25 millionc 3,264–3,648d 6,201c 55,411c 0.1% in >85% of base callsc 45.6 × 48 × 51.8c Illumina 2 × 75c 3.3–3.8c 13.6 13c 44 million–50 millionc 1,632–1,879d 6,201c 55,411c 0.1% in >85% of base callsc 45.6 × 48 × 51.8c Illumina 2 × 150c 6.6–7.5c 26.9 24c 44 million–50 millionc 827–940d 6,201c 55,411c 0.1% in >80% of base callsc 45.6 × 48 × 51.8c Illumina Illumina MiSeq Reagent kit v2 1 × 36c 0.54–0.61c 2.2 4c 12 million–15 millionc 7,946–8,976d 4,847c 108,244c 0.1% in >90% of base callsc 68.6 × 56.5 × 52.3c Illumina 2 × 25c 0.75–0.85c 3.1 5.5c 24 million–30 millionc 5,702–6,463d 4,847c 108,244c 0.1% in >90% of base callsc 68.6 × 56.5 × 52.3c Illumina 2 × 150c 4.5–5.1c 18.3 24c 24 million–30 millionc 950–1,077d 4,847c 108,244c 0.1% in >80% of base callsc 68.6 × 56.5 × 52.3c Illumina 2 × 250c 7.5–8.5c 30.5 39c 24 million–30 millionc 570–646d 4,847c 108,244c 0.1% in >75% of base callsc 68.6 × 56.5 × 52.3c Illumina Illumina MiSeq Reagent kit v3 2 × 75c 3.3–3.8c 13.6 21c 44 million–50 millionc 1,362–1,568d 5,174c 108,244c 0.1% in >85% of base callsc 68.6 × 56.5 × 52.3c Illumina 2 × 300c 13.2–15c 53.8 56c 44 million–50 millionc 345–392d 5,174c 108,244c 0.1% in >70% of base callsc 68.6 × 56.5 × 52.3c Illumina Illumina NextSeq 500 Mid Output 2 × 75c 16.3–20c 71.8 15c <260 millionc 318–391d 6,369c 266,835c 0.1% in >75% of base callsc 53.3 × 63.5 × 58.4c Illumina 2 × 150c 32.5–39c 140 26c <260 millionc 163–196d 6,369c 266,835c 0.1% in >80% of base callsc 53.3 × 63.5 × 58.4c Illumina Illumina NextSeq 500 High Output 1 × 75c 25–30c 107.7 11c <400 millionc 312–374d 9,347c 266,835c 0.1% in >80% of base callsc 53.3 × 63.5 × 58.4c Illumina 2 × 75c 50–60c 215.3 18c <800 millionc 156–187d 9,347c 266,835c 0.1% in >80% of base callsc 53.3 × 63.5 × 58.4c Illumina 2 × 150c 100–120c 430.6 29c <800 millionc 78–93d 9,347c 266,835c 0.1% in >75% of base callsc 53.3 × 63.5 × 58.4c Illumina Single-molecule real-time sequencing Pacific Biosciences RS II P6-C4 chemistry >20,000c 8–16e 57.4 0.5–4c 55,000c 250–500d 4,000e 695,000 14% errors per base 203.0 × 90.0 × 160.0c PacBio, AllSeqf (89) Pacific Biosciences Sequel system >20,000c 80–160e 574.2 0.5–6c 370,000c 70–140d 11,200e 350,000 14% errors per base 92.7 × 86.4 × 167.6c PacBio, AllSeqf (89) Oxford Nanopore MinION Mk1 (1D) >882,000 10–20c 71.8 1.67–72c 138,000 49.95–99.9d 999c 1,000c 12% errors per base 10.5 × 3.3 × 2.3c Oxford Nanopore Technologies, Loman Labsg (231) Oxford Nanopore MinION Mk1 (2D) >882,000 10–20c 71.8 1.67–72c 138,000 49.95–99.9d 999c 1,000c 15% errors per base 10.5 × 3.3 × 2.3c Oxford Nanopore Technologies, Loman Labsg (231, 232) Oxford Nanopore PromethION single flow cell <300,000c 233c 836.2 1.67–>72c 26 millionc NA NA 135,000 (PEAP)c NA 44.0 × 24.0 × 40.0c Oxford Nanopore Technologies Oxford Nanopore PromethION 48 flow cells <300,000c 11,000c 3,9475.8 1.67–>72c 1.25 billionc NA NA 135,000 (PEAP)c NA 44.0 × 24.0 × 40.0c Oxford Nanopore Technologies ↵a All quantitative performance measures were taken from previously reported data, as indicated. Consumables costs were calculated as follows: Illumina costs included PhiX Control kit v3, the Nextera XT DNA sample preparation kit (96 samples)/Nextera DNA library preparation kit (96 samples), and Nextera XT Index kit v2 (96 indexes and 384 samples), the highest-output reagent kit. PEAP, PromethION Early-Access Program; NA, no data available.
↵b Calculated for 96 samples and the genome size of S. aureus strain MRSA252 (2,902,619 bp).
↵c Manufacturer's data.
↵d Estimated calculation for consumables.
↵e For 16 SMRT cells.
↵f See http://www.allseq.com/knowledge-bank/sequencing-platforms/pacific-biosciences/.
↵g See http://lab.loman.net/2017/03/09/ultrareads-for-nanopore/.
- TABLE 2
Performance analysis of assembly toolsa
Analysis tool (reference[s]) Concept Computational requirement Speed Assembly quality Preferred sequencing technology(ies) Web address(es) Input format Output format(s) Web based Velvet (103, 126) de Bruijn graph-based assembly that resolves repeat-rich regions; can be used for de novo or reference-guided assembly; requires paired reads with 20- to 25-fold coverage Mid* Medium* Low* Illumina https://cge.cbs.dtu.dk/services/Assembler/ FASTA, FASTQ, SAM, or BAM AMOS, modified FASTA SPAdes/hybridSPAdes (112) de Bruijn graph-based assembler for de novo assembly of short and long reads Low** Low** Mid*/** Mixed input (Illumina, Ion Torrent, PacBio CLR, Oxford Nanopore) https://cge.cbs.dtu.dk/services/SPAdes/ FASTA, FASTQ, or BAM FASTA, FASTQ, FASTG Command line IDBA-UD (108) de Bruijn graph-based assembly designed for assembly of repeat-rich reads of various sequencing depths Low* Medium* Mid* Illumina http://i.cs.hku.hk/~alse/hkubrg/projects/idba_ud/ FASTA FASTA RAY (96) de Bruijn graph-based assembly that uses seeds instead of Eulerian walks; used for de novo assembly; designed for short reads Low*** Fast*** Low*** Mixed input (454, Illumina, Ion Torrent) http://denovoassembler.sourceforge.net/ FASTA, FASTQ, or SFF FASTA, TXT Minimap/miniasm (116) OLC framework that computes overlaps and performs read trims and unitig construction; can be used for de novo or reference-guided assembly Low** High** High*/** PacBio, Oxford Nanopore https://github.com/lh3/minimap, https://github.com/lh3/miniasm FASTA GFA, PAF Canu (118) OLC framework that computes overlaps and performs read correction, read trims, and unitig construction; used for de novo assembly Mid** Low** High*/** PacBio, Oxford Nanopore https://github.com/marbl/canu FASTA or FASTQ FASTA ↵a All quantitative performance measures were taken from data reported previously, as indicated. CLR, continuous long reads; GFA, graphical fragment assembly; PAF, pairwise mapping format; SFF, standard flowgram format (454 data format); *, E. coli K-12 MG1655 data set (110); **, Enterobacter kobei data set (233); ***, Illumina data from E. coli (SRA accession number SRX000429) (234). Note that for SPAdes, only the nonhybrid tool is accessible as a Web-based tool.
- TABLE 3
Overview of genome characterization toolsa
Analysis tool (reference[s]) Concept(s) Input type(s) Input format(s) Output format(s) Web address Identification Web based KmerFinder (121, 122) Uses k-mers to identify strain using WGS data Raw sequences, contigs FASTQ, FASTA Tab delimited, online https://cge.cbs.dtu.dk/services/KmerFinder/ NCBI BLASTb (123) NCBI Web-based interface for performing BLAST searches; searches hits in the database that match the given sequence Contigs FASTA Online, tab delimited https://blast.ncbi.nlm.nih.gov/Blast.cgi MLST Web server (125) Web-based database that identifies STs from short sequencing reads or draft genomes Raw sequences, contigs FASTQ, FASTA Online https://cge.cbs.dtu.dk/services/MLST/ Command line PathoScope 2.0 (127) Complete framework based on Bayesian missing-data approach, for direct strain identification Raw sequences FASTQ, FASTA Tab delimited https://sourceforge.net/p/pathoscope/wiki/Home/ Annotation Web based RAST (129) Web-based server for localization and identification of tRNA, rRNA, and coding sequences; includes a browser for screening the output Contigs FASTA GenBank, EMBL, GFF3, GTF, Excel, and tab delimited http://rast.nmpdr.org/ Command line PROKKA (132) Rapid annotation tool for localization and identification of rRNA, tRNA, tmRNA, signal peptides, noncoding RNA, and coding sequences Contigs FASTA FASTA, tab delimited, SQN, GenBank file, GFF3 http://www.vicbioinformatics.com/software.prokka.shtml Virulence Web based VirulenceFinder Detect virulence genes in WGS data using the BLAST algorithm Raw sequences, contigs FASTQ, FASTA Tab-delimited summary, FASTA https://cge.cbs.dtu.dk/services/VirulenceFinder/ VFDB (138) Source of virulence information, including Web-based service to perform BLAST to detect virulence genes Contigs FASTA Online, tab delimited http://www.mgc.ac.cn/VFs/ Antimicrobial resistance Web based ResFinder Detects resistance genes in WGS data Raw sequences, contigs FASTQ, FASTA Tab-delimited summary, FASTA https://cge.cbs.dtu.dk/services/ResFinder/ RGI/CARD (144–146) Web-based as well as command line versions available to perform resistance gene detection using the CARD database Contigs, GenBank accession no. FASTA, GenBank accession no. (nucleotide or protein) JSON, tab-delimited summary, FASTA, heat map PDF https://card.mcmaster.ca/analyze/rgi PlasmidFinder Tool to detect plasmids in WGS data Raw sequences, contigs FASTQ, FASTA Tab-delimited summary, FASTA https://cge.cbs.dtu.dk/services/PlasmidFinder/ CGE BAP (107) Web-based suite for automated genomic characterization; if raw sequence reads are provided, performs assembly; set of tools is applied to the contigs, ResFinder, VirulenceFinder, and PlasmidFinder Raw sequences, contigs FASTQ, FASTA Tab-delimited summaries, FASTA https://cge.cbs.dtu.dk/services/cge/ - TABLE 4
Performance analysis of comparative genomics toolsa
Analysis tool (reference[s]) Concept Method Run time (h) Topology score (%) Web address(es) Input type(s) Input format(s) Output format(s) Web based PubMLST (158) Web-accessible database where it is possible to run cgMLST and wgMLST analyses cgMLST/wgMLST NA NA https://pubmlst.org/ Contigs FASTA cgMLST/wgMLST profile CSI Phylogeny 1.4 (161) High-quality SNP method using reference mapping of reads and mapping and SNP calling assessments Reference-based SNP ND ND https://cge.cbs.dtu.dk/services/CSIPhylogeny/ Raw sequences, contigs FASTA, FASTQ ND NDtree 1.2 (161) Creates k-mers of reads and maps them to a reference; performs simple model to determine no. of SNPs Statistical method 3–3.5b ND https://cge.cbs.dtu.dk/services/NDtree/ Raw sequences FASTQ Newick Command line kSNP3 (154, 155) Uses k-mer analyses to detect SNPs between strains without using either multiple-sequence alignment or a reference genome Non-reference-based SNP 0.5c 91.80–95.80c,e https://sourceforge.net/projects/ksnp/ Raw sequences, contigs FASTA Newick, MSA Roary (169) Tool for constructing pangenomes from contigs Pangenome 4.30d 100d https://sanger-pathogens.github.io/Roary/ Contigs GFF3 FASTA, TXT, CSV, Rtab Pan-Seqf (175) Pangenome assembler with additional locus finder for core/accessory gene allele profiles (a Web-based version is also available) Pangenome ND ND https://github.com/chadlaing/Panseq, https://lfz.corefacility.ca/panseq/ Contigs FASTA TXT, FASTA Lyve-SET (179) High-quality SNP method using reference mapping of reads and mapping and SNP calling assessments Reference-based SNP 6.25c 85c https://github.com/lskatz/lyve-SET Raw sequences, contigsg FASTA, FASTQ Matrix, FASTA, Newick, VCF SPANDx (182) Complete workflow for creating SNP/indel matrixes as well as locus presence/absence matrixes from raw sequencing reads from a range of NGS technologies Reference-based SNP 3.1c 100c https://sourceforge.net/projects/spandx/ Raw sequences FASTA, FASTQ NEXUS ↵a All quantitative performance measures were taken from previously reported data, as indicated. ND, no data; NA, not applicable; MSA, multiple-sequence alignment; GFF3, General Feature Format 3; VCF, variant call format.
↵b Based on 46 VTEC genomes (20).
↵c Based on 21 E. coli genomes (167).
↵d Wall time for 1,000 S. enterica serovar Typhi genomes (169).
↵e Using core.
↵f A Web-based version is also available.
↵g Contigs are simulated to reads.
- TABLE 5
Performance analysis of phylogeny toolsa
Command line analysis tool (reference) Concept Run time (h) Accuracy (%) Input format Output format RAxML (191) Maximum likelihood phylogenetic tree estimator tool; slow but very accurate 612b 84.47c PHYLIP or FASTA Newick FastTree (163) Approximately maximum likelihood phylogenetic tree estimator; fast but slightly less accurate 2.63b 83.6c PHYLIP or FASTA Newick MrBayes (198) Bayesian-based phylogenetic tree; complex to define models and not user-friendly ND ND NEXUS NEXUS ↵a All quantitative performance measures were taken from previously reported data, as indicated. The input type for all of these tools is aligned reads/SNPs. ND, no data.
↵b Averages for 3 large biological data sets aligned via 3 different methods (TrueAln, PartTree, and Quicktree) (197).
↵c Accuracy = 100% − missing branch rates (%) for 3 large biological data sets aligned via 3 different methods (TrueAln, PartTree, and Quicktree) (197).
- TABLE 6
Overview of complete analysis software suitesa
Software suite Concept RAM compatibility (Gb) Run time (h) No. of schemes Price ($) Source or Web address Input format Output format(s) Commercial BioNumerics 7.6.2 Suite containing multiple modules, thereby having many functionalities; able to perform wgMLST ND ND 14 Request quoteb Applied Maths FASTQ Depends on module Ridom SeqSphere+ Suite dedicated to outbreak analyses; customizable automation flows for processing raw reads to phylogeny using either cgMLST or wgMLST; for cgMLST, it includes CT definitions 16–32c ND 7 2,500d Ridom Bioinformatics FASTQ CT, phylogeny Free NCBI Pathogen Detection (beta) NCBI-provided Web service with main focus on detection of foodborne pathogens; automated flow from raw sequences to phylogeny inference NA NA 19e Free https://www.ncbi.nlm.nih.gov/pathogens/ FASTQ Web-accessible SNP tree, AMR data - TABLE 7
Pros and cons of sequencing platforms
Platform Pros Cons Sequencing by synthesis Illumina Technology used widely by the WGS industry; lowest per-Gb sequencing cost range; highest confirmed output; wide range of Illumina machines suited for a wealth of applications and demands; lowest error rates Rehybridization of template strands and low-copy-no. yields during bridge amplification; use of potentially biased DNA polymerases during bridge amplification; incomplete base extension (phasing, prephasing); shortest read lengths; long sequence runs; high instrument costs; no real-time data access Single-molecule real-time sequencing Pacific Biosciences Fast sequence runs; long reads suitable for assembly of draft genomes and completion of genome assemblies; possibility of obtaining epigenetic sequence information; real-time measurement of base incorporation Possibility of false detection of unincorporated nucleotides during sequencing; largest instrument footprint; low output per run; high error rates Oxford Nanopore Technologies Fast sequencing; longest confirmed reads; smallest instrument footprint; lowest instrument and consumables costs; real-time measurement of base incorporation; real-time data output Sensitivity of biological nanopores to changes in exptl environment; highest error rate of all platforms; the performance of the PromethION machine is not experimentally validated - TABLE 8
Pros and cons of analysis tools
Algorithm Interface type(s) Pro(s) Con(s) Assembly Velvet Web based Designed for repeat-rich reads; automated parameter tuning for quality control; detailed tutorial; Web-based accessibility Small N50 contig size; technology specific; coverage cutoff excludes potentially correct low-coverage vertices; high memory usage; suitable for short reads only IDBA-UD Command line Designed for repeat-rich short reads with various sequencing depths; among the lowest memory usages; error correction after each iteration for quality control Technology specific; no tutorial; suitable for short reads only RAY Command line Hybrid assembly of multiple sequencing platform reads; heuristics for contig length determination that increase quality of sequence accuracy; automated parameter calculation; detailed tutorial Small N50 contig size; poor performance with lower-quality reads; suitable for short reads only SPAdes/hybridSPAdes Web based Hybrid assembly of multiple sequencing platform reads; suited for short and long reads; among the lowest memory usages; largest N50 contig size; closing of gaps and resolution of repeats in assembly graph for quality control; option to merge contigs from other assemblers; detailed tutorial; Web-based accessibility Longest computing time Minimap/miniasm Command line Shortest computing time; compatibility with other overlapping workflows when converted to PAF format; detailed tutorial Technology specific; no sequencing error correction; missing overlaps and misassemblies during graph cleaning; suitable for long reads only Canu Command line Large N50 contig size; detailed tutorial; initial read correction to remove noise for quality control Long computing time; high memory usage; suitable for long reads only Genome characterization Identification KmerFinder Web based No bioinformatics skills required; easy to use; easy to interpret output; raw sequence or contig input; possible to detect contamination Method should be set properly; no assembly is performed NCBI BLAST Web based Largest database; multiple databases; multiple tools available Interpretation of results can be difficult; some BLAST knowledge is advised MLST Web server Web based Simple online workflow; no bioinformatics skills required Suitable for samples of single species only; accepts short reads only from Illumina, Roche 454, Ion Torrent, and SOLiD PathoScope 2.0 Command line Able to detect contamination; quality control of raw sequencing reads; complete workflow that minimizes the need for intense computational background; detailed and understandable tutorial When testing samples with multiple strains of one species, parsimony can lead to missing of strains due to reassignment; for nearly identical strains, a coverage of >20% is necessary to distinguish between them; long computing time Annotation RAST Web based Web accessible; KEGG connection; graph presentation Long waiting times; must send data to server PROKKA Command line Short computing time; parallel annotation with 5 tools in a single workflow; detailed tutorial Decreased annotation performance with understudied or draft genomes; suitable only for samples of single species Virulence VirulenceFinder Web based Easy to use; fast results; parameter control; raw sequence or contig input Not able to detect SNP-related virulence; available for only limited groups of species/genera VFDB Web based Extended wealth of information; more markers associated with virulence than in VirulenceFinder Function to detect virulence markers is not easy to use; not able to detect SNP-related virulence AMR ResFinder Web based Fast results; parameter control; raw sequence or contig input Not able to detect SNP-related resistance; not able to detect ampC RGI/CARD Web based Able to detect SNP-related resistance; accession no. input possible; raw sequence or contig input; access to antibiotic resistance ontology; BLAST present; graphical views Limited contig upload size (<20 Mb); no raw sequence data input possible PlasmidFinder Web based Raw sequence or contig input Limited database; detects only plasmids and does not include the presence of AMR CGE BAP Web based Complete suite for genome characterization; easy to use Need for subscription for access; long computing times; no annotation performed Comparative genomics PubMLST Web based Creates source for both MLST and cgMLST as other sets of genes used for typing; built on BIGSdb, which makes it locally installable; all databases can be downloaded; user is able to contribute to the database Finding correct data can be difficult; built to share data publically CSI Phylogeny 1.4 Web based Raw read and contig input possible; hqSNPs by selecting SNPs based on strict criteria; many parameters can be set Only reference-based comparison; need to provide reference sequence; amt of parameters could be confusing for clinician without bioinformatics knowledge NDtree 1.2 Web based Raw read input, which makes it able to skip assembly; easy to use; automatic selection of best reference using KmerFinder Method is not comparable to others; fixed parameters; lack of documentation; only reference-based comparison kSNP3 Command line Very fast method; automatically skips regions with high mutation frequency; easily scalable; all-to-all comparison possible; works with raw sequence data and/or contigs as input Compared to other comparative genomics tools, overall accuracy is slightly low; no hqSNP method; bioinformatics knowledge needed Roary Command line Protein misprediction control; detailed manual; construction of pangenome Input has to be contigs; slow computation with larger sample sizes; relies fully on annotation accuracy Pan-Seq Command line and Web based Minimal user interaction needed; construction of pangenome Input has to be contigs; no exptl data on computing speed and accuracy Lyve-SET Command line Extensive SNP filtering (hqSNP); implementation for running on a computing cluster is present Can be too conservative in SNP calling; only reference-based comparison; bioinformatics knowledge needed SPANDx Command line Extensive error checking, filtering, and variant identification steps during quality control (hqSNP); complete workflow from raw reads to comparative analysis; quick variant visualization through automatically generated presence/absence matrixes and error-corrected SNP and indel matrixes; works with raw sequence data as input Only reference-based comparison; bioinformatics knowhow needed Phylogeny RAxML Command line Enables standard nonparametric bootstrapping, rapid bootstrapping, bootstopping, and calculation of SH-like support values for quality control; CAT and Shimodaira-Hasegawa test for quality control; comprehensive workflow; detailed manual; GTR model available Longest computing time; highest accuracy; computationally expensive FastTree Command line Shortest computing time; CAT and Shimodaira-Hasegawa test for quality control; GTR model available; detailed manual Lowest accuracy due to limited initial tree improvement MrBayes Command line Possible to optimize a model; most models available for all phylogeny methods; detailed manual; GTR model available Input and output formats in NEXUS; complex to use Complete outbreak analysis software suites BioNumerics 7.6.2 Local suite Easy to use; custom schemes possible; scheme modification; wgMLST; cgMLST; rMLST; most schemes present Separate modules needed; no cluster types Ridom SeqSphere+ Local suite Easy to use; use of cluster types; ad hoc schemes possible; cgMLST; wgMLST Database can be slow with many samples; fewer schemes available than with BioNumerics NCBI Pathogen Detection (beta) Web-based suite Free to use; direct link to foodborne pathogen outbreaks; data sharing; uses collection of strains Registration needed; focus on foodborne pathogens; data are publically available; time-consuming to register new samples; not suitable for real-time hospital-acquired outbreaks
