2013 NGS Field Guide – Table 1 – Instrument Use Grades

Table 1a. Utility of 2nd and 3rd generation DNA sequencing platforms for de novo assemblies of different templates. Table assumes Illumina, Ion Torrent, and Oxford Nanopore achieve stated goals (independent of stated time-lines). Initial letter indicates the author’s opinion of the overall utility (grade) for a platform for a specific application. Utility grades combine data characteristics (amount, quality, length), cost of data, and ease of assembling the data into the final desired product. Major considerations for utility grades are noted.

Platform – instrument Application: de novo assemblies
BACs, plastids, & microbial genomes transcriptome Plant & animal genome
454 – GS Jr. B – good but expensive C – need multiple runs, expensive D – cost prohibitive
454 – FLX+ A – good, need to multiplex to be economical B – good but expensive, libraries usually normalized, not best for short RNAs C – OK as part of a mixed platform strategy, prohibitive to use alone
MiSeq – v2 A – good, need to multiplex for best economics A/B –expensive for rare transcripts (compared to HiSeq), but reads are longer for better assembly B – expensive relative to HiSeq, but additional read length can be valuable
HiSeq 2000/2500, standard run B/C – more data than needed unless highly indexed; assembly more challenging than 454 or MiSeq A – good, assembly more challenging than 454 but much more data available for analyses A – primary data type in many current projects; requires mate-pair libraries
HiSeq 2500, rapid run (projected) B – more data than needed unless highly indexed; assembly more challenging than 454 A – good, assembly more challenging than 454 but much more data available for analyses A – will probably be more expensive than HiSeq2000, but increased read length may be worth it
Ion Torrent – 314 B/C – OK, lowest experimental cost but reads are shorter & more expensive than Illumina C – OK, but reads are shorter & more expensive than Illumina D – cost prohibitive, reads shorter than alternatives
Ion Torrent – 318 B/A – good, less data than MiSeq B/A – good, less data than MiSeq, reads similar to 454 titanium but less expensive C – high cost relative to Proton or Illumina, more economical than 454 for mixed platform strategy
Ion Torrent Proton I B – more data than needed unless indexed; assembly more challenging than 454 or Illumina B/A – assembly currently more challenging than Illumina or 454 B – expensive relative to HiSeq or Proton II/III
Ion Torrent Proton II (projected) B/C – more data than needed unless highly indexed; assembly more challenging than 454 or Illumina B/A – assembly currently more challenging than Illumina or 454 A/B – should be similar to HiSeq
Ion Torrent Proton III (forecast) C – more data than needed unless highly indexed B/A – need assembly pipelines A – cost per MB could make it the best
minION (forecast) B – more expensive than GridION A/B – more expensive than GridION C/D – more expensive than GridION
GridION (forecast) B/A – great for scaffolding, will need to combine with short reads with lower error rate until error rates are reduced B/A – great for defining full-length transcripts, will need to combine with short reads with lower error rate until error rates are reduced B/A – great for scaffolding, will need to combine with short reads with lower error rate until error rates are reduced
SOLiD – 5500 C – more data than needed unless highly indexed; assembly more challenging than 454 or Illumina C/D – short reads make assembly challenging or impossible C/D – short reads make assembly challenging or impossible
PacBio – RS B – good for hybrid assemblies; not economical for solo assemblies – requires high coverage due to high error rates B/D – good for hybrid assemblies; too expensive for solo use; short RNA is challenging B/D – good for hybrid assemblies & scaffolding (mixed platform strategy); cost prohibitive for solo use

Table 1b. Utility of 2nd and 3rd generation DNA sequencing platforms for resequencing applications. Table assumes HiSeq2500, Ion Torrent Proton, and Oxford Nanopore achieve stated goals (independent of stated time-lines). Initial letter indicates the author’s opinion of the overall utility (grade) for a platform for a specific application. Utility grades combine data characteristics (amount, quality, length), cost of data, and ease of assembling the data into the final desired product. Major considerations for utility grades are noted.

Platform – instrument Application: resequencing
Targeted loci Transcript counting Genome resequencing
454 – GS Jr. B/C – good but expensive, need to limit loci D – cost prohibitive D – cost prohibitive for large genomes
454 – FLX+ B – good but expensive, should limit loci D – cost prohibitive D – cost prohibitive for large genomes
MiSeq A/B – good, fewer and higher cost reads than HiSeq B – more expensive than HiSeq or SOLiD or ProtonII+ B/C – expensive for large genomes
HiSeq 2000/2500 – standard run A – primary data type in many current projects; best for many loci A – primary data type in many current projects A – primary data type in many current projects
HiSeq 2500 – rapid run (projected) A – faster path to leading data type A/B – likely to be slightly more expensive than with standard flow cell A – faster path to leading data type
Ion Torrent – 314 C – OK but expensive, need to limit loci D – cost prohibitive D – cost prohibitive
Ion Torrent – 318 B – good, slightly less data per run than MiSeq B/C – more expensive than HiSeq or SOLiD; new informatics pipelines needed; new error profile C – expensive for large genomes
Ion Torrent Proton I A/B – similar to MiSeq, but different error profile will inhibit switching B – more expensive than Illumina or SOLiD; new informatics pipelines needed (different error profile than Illumina) B – expensive relative to HiSeq or Proton II+
Ion Torrent Proton II (projected) A/B – similar to HiSeq, but different error profile will inhibit switching A/B – new informatics pipelines needed A – supposed to set new pricing standard, could become leading shorter-read platform
Ion Torrent Proton III (forecast) A/B – costs projected to be better than HiSeq; error profile different than Illumina A/B – new informatics pipelines needed A – supposed to set new pricing standard, could become leading shorter-read platform
minION C/D? – error profile may make this less desirable D – probably cost prohibitive C/D? – expensive for large genomes
GridION B? – error profile may make this less desirable B? – error profile may make this less desirable B? – error profile may make this less desirable
SOLiD – 5500xl B – harder to assemble than Illumina A/B – used much less than HiSeq A/B – used much less than HiSeq
PacBio – RS C/D – expensive but can sequence difficult regions D – cost prohibitive C/D – cost prohibitive except for strutural variants

Table 1c. Utility of 2nd and 3rd generation DNA sequencing platforms for various applications. Table assumes HiSeq2500, Ion Torrent Proton, and Oxford Nanopore achieve stated goals (independent of stated time-lines). Initial letter indicates the author’s opinion of the overall utility (grade) for a platform for a specific application. Utility grades combine data characteristics (amount, quality, length), cost of data, and ease of assembling the data into the final desired product. Major considerations for utility grades are noted.

Platform – instrument Various Applications
Metagenomics1 Mutation Detection2 Other limitations
454 – GS Jr. B/C – good but costs limit sample number & depth D – cost prohibitive Customer happiness
454 – FLX+ B – good but expensive, long reads maximize data per read C – expensive, good for identifying clusters Reliability issues; customer happiness
MiSeq – v2 A/B – good; shorter reads than 454, but much greater depth; beware phase* B – more expensive than HiSeq, SOLiD, or Proton II+ No inexpensive kits
HiSeq 2000/2500 – standard run B – OK, limited by short reads, beware phase* A – primary data type in many current projects Increasing reagent costs
HiSeq 2500 – rapid run (projected) A/B – good, limited by short reads, beware phase* A – primary data type in many current projects What will reagents cost?
Ion Torrent – 314 B/C – limited by short reads and cost D – cost prohibitive Escalating costs with increased read length
Ion Torrent – 318 A/B – 400 base reads could allow this to replace 454; no phase issues, unlike MiSeq B – more expensive than MiSeq, HiSeq, SOLiD, or Proton Escalating costs with increased read length
Ion Torrent Proton I B – shorter read-length than 454 or MiSeq, longer reads than HiSeq B – more expensive than HiSeq or SOLiD; different error profile Will 400 base reads become available? Reductions in kit costs?
Ion Torrent Proton II (projected) A/B – shorter read-length than 454 or MiSeq, longer reads than HiSeq A/B – similar in cost to HiSeq & SOLiD; different error profile ≥400 base reads? Will they finally catch Illumina?
Ion Torrent Proton III (forecast) A/B – shorter read-length than 454 or MiSeq, longer reads than HiSeq B – projected to be cost leader; different error profile ≥400 base reads? Will they finally beat Illumina before the nanopores come on-line?
minION (forecast) B – excellent for environmental sample sequencing; enrichment techniques will need to be developed; field portable; limited by accuracy D? – accuracy is likely limiting for this application When will this be available? When will data be publicly available for non-commercial software development?
GridION (forecast) A/B – excellent for environmental sample sequencing; enrichment techniques will be helpful; limited by accuracy ? – accuracy is likely limiting for this application When will this be available?  How much will the nodes and reagents cost?
SOLiD – 5500xl D – limited by short reads B – frequent data type in many current projects Limited future
PacBio – RS C/D – high costs, but long reads allow scaffolding in hybrid sequencing strategies F – cost prohibitive & insufficient accuracy except for short consensus sequence reads When will new applications be available?

1Metagenomics – characterization of 16S sequences within and among microbial communities, primarily via sequencing amplicons.

2Mutation Detection – identification of rare sequence variants

*Illumina instruments require a mixture of different base signals among clusters during each cycle, thus amplicon sequencing requires strategies to offset the beginning bases of amplicons or use of custom sequencing primers

  • Henry

    Please re-confirm on the table:
    Application: de novo assemblies and Application: resequencing
    Did these 2 mixed up? I think I cannot highly multiplex de-novo assemblies in MiSeq (or even ion-photon) , unless they are really small genomes like bacteria.

  • Travis

    The table is correct – you will see that there are separate de novo columns for very small genomes (BACs, plastids, microbes), then transcriptomes, then plant & animal genomes (i.e., going from smallest to largest). Of course there is a lot of variation in microbial (as well as plant & animal) genomes, but this gives a general indication of scale & application. We have specific spreadsheets that can help do the calculations quite precisely. Some great graduate students are working on reformatting, documentation & even creating a web server. Hopefully we can release those to the community soon.