Review doi:10.

1038/nature24286

DNA sequencing at 40: past, present

and future
Jay Shendure1,2, Shankar Balasubramanian3,4, George M. Church5, Walter Gilbert6, Jane Rogers7, Jeffery A. Schloss8 &
Robert H. Waterston1

This review commemorates the 40th anniversary of DNA sequencing, a period in which we have already witnessed
multiple technological revolutions and a growth in scale from a few kilobases to the first human genome, and now to
millions of human and a myriad of other genomes. DNA sequencing has been extensively and creatively repurposed,
including as a ‘counter’ for a vast range of molecular phenomena. We predict that in the long view of history, the impact
of DNA sequencing will be on a par with that of the microscope.

D
NA sequencing has two intertwined histories—that of the under- v­ isualization in two dimensions, with the resulting positions diagnostic
lying technologies and that of the breadth of problems for which of their size and sequence5.
it has proven useful. Here we first review major developments
in the history of DNA sequencing technologies (Fig. 1). Next we consider The invention of DNA sequencing
the trajectory of DNA sequencing applications (Fig. 2). Finally, we discuss Early attempts to sequence DNA were cumbersome. In 1968, Wu reported
the future of DNA sequencing. the use of primer extension methods to determine 12 bases of the cohesive
ends of bacteriophage lambda6. In 1973, Gilbert and Maxam reported 24
History of DNA sequencing technologies bases of the lactose-repressor binding site, by copying it into RNA and
The development of DNA sequencing technologies has a rich history, sequencing those fragments. This took two years: one base per month7.
with multiple paradigm shifts occurring within a few decades. Below, we The development, in around 1976, of two methods that could decode
review early efforts to sequence biopolymers, the invention of electro- hundreds of bases in an afternoon transformed the field. Both methods—
phoretic methods for DNA sequencing and their scaling to the Human the chain terminator procedure developed by Sanger and Coulson, and
Genome Project, and the emergence of second (massively parallel) and the chemical cleavage procedure developed by Maxam and Gilbert—used
third (real-time, single-molecule) generation DNA sequencing. Some key distances along a DNA molecule from a radioactive label to positions
technical milestones are also summarized in Box 1. occupied by each base to determine nucleotide order. Sanger’s method
involved four extensions of a labelled primer by DNA polymerase, each
Early sequencing with trace amounts of one chain-terminating nucleotide, to produce frag-
Fred Sanger devoted his scientific life to the determination of p ­ rimary ments of different lengths8. Gilbert’s method took a terminally labelled
sequence, believing that knowledge of the specific chemical structure DNA-restriction fragment, and, in four reactions, used chemicals to create
of biological molecules was necessary for a deeper understanding1. base-specific p­ artial cleavages9. For both methods, the sizes of fragments
Ironically, given the state of sequencing technology for each biopolymer present in each base-specific reaction were measured by electrophoresis
today, proteins and RNA came first. on polyacrylamide slab gels10, which enabled separation of the DNA
The first protein sequence, of insulin, was determined in the early 1950s fragments by size with single-base resolution. The gels, with one lane per
by Sanger, who fragmented its two chains, deciphered each fragment and base, were put onto X-ray film, producing a ladder image from which the
overlapped the fragments to yield a complete sequence. His work showed sequence could be read off immediately, going up the four lanes by size
unequivocally that proteins had defined patterns of amino acid residues2. to infer the order of bases.
The later development of Edman degradation, a repeated elimination of These methods came into immediate use. Shotgun sequencing—
an N-terminal residue from the peptide chain, made protein sequencing sequencing of random clones followed by sequence assembly based on
easier3. Although these methods were cumbersome, many proteins had the overlaps—was suggested by Staden in 197911, greatly facilitated by
been sequenced by the late 1960s, and it became clear that each protein’s Messing’s development of the single-stranded M13 phage cloning vector
sequence varied across species and between individuals. around 198012, and used to assemble genomes de novo, such as bacterio-
In the 1960s, RNA sequencing was tackled by this same general process: phage lambda as early as 198213. By 1987, automated, fluorescence-based
an RNA species was first fragmented with RNases, next the pieces were Sanger sequencing machines, developed by Smith, Hood and Applied
separated by chromatography and electrophoresis, then individual frag- Biosystems14,15, could generate around 1,000 bases per day. Sequence
ments were deciphered by sequential exonuclease digestion, and finally data grew exponentially, approximating Moore’s law and motivating the
the sequence was deduced from the overlaps. The first RNA sequence, of creation of central data repositories (such as GenBank) that, through
alanine tRNA, required five people working three years with one gram of search tools (such as BLAST16), amplified the value of each sequence and
pure material (isolated from 140 kg of yeast) to determine 76 nucleotides4. engendered a spirit of data sharing. By 1982, over half a million bases had
This process was greatly simplified by ‘fingerprinting’ techniques, which been deposited in GenBank; by 1986, nearly 10 million bases (GenBank
included the separation of radioactively labelled RNA fragments and and WGS Statistics; https://www.ncbi.nlm.nih.gov/genbank/statistics/).

1
Department of Genome Sciences, University of Washington, Seattle, Washington, USA. 2Howard Hughes Medical Institute, Seattle, Washington, USA. 3Department of Chemistry, University of
Cambridge, Cambridge, UK. 4Cancer Research UK Cambridge Institute, University of Cambridge, Cambridge, UK. 5The Wyss Institute & Department of Genetics, Harvard Medical School, Boston,
Massachusetts, USA. 6Department of Molecular and Cellular Biology, Harvard University, Cambridge Massachusetts, USA. 7International Wheat Genome Sequencing Consortium, Little Eversden,
Cambridge, UK. 8National Human Genome Research Institute, National Institutes of Health, Bethesda, Maryland, USA.

1 9 o c t o b e r 2 0 1 7 | V O L 5 5 0 | N A T U RE | 3 4 5
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
 RESEARCH Review

First generation sequencing (Sanger)

sequencing; (6) capillary electrophoresis, which eliminated the pouring
1 Genomic DNA
and loading of gels, while also simplifying the extraction and interpretation
2 Fragmented
of the fluorescent signal21; (7) adoption of industrial processes to maximize
DNA efficiencies and minimize errors (for example, automation, quality control,
3 Cloning and amplification standard operating procedures, and so on).
Wet laboratory protocols were only half the challenge. Substantial effort
3′... * $ & 7 $ * $ 7 & & * $ * & * 7 * $ ...5′
4 Sequencing was invested into the development of software to track clones, and into
5′... & 7 * $ 7 ... ...& 7 * $ 7 & ... the interpretation and assembly of sequence data. For example, manual
...& 7 * $ 7 & 7 ... editing of the sequence reads was replaced by the development of phred,
...& 7 * $ 7 & 7 $ ...
5 Detection which introduced reliable quality metrics for base calls and helped sort
...& 7 * $ 7 & 7 $ * ...
&7*$7&7$**&7&*&$&7 out closely related repeat sequences22. Individual reads were then assem-

...
...& 7 * $ 7 & 7 $ * * & 7 & * & $ & 7 ... bled from overlaps in a quality-aware fashion to generate long, continuous
Second generation sequencing (massively parallel) stretches of sequence. As more complex genomes were tackled, ­repetitive
1 Genomic DNA
sequences were increasingly confounding. Even after deep shotgun
sequencing of a BAC, some sequences were not represented, resulting
2 Fragmented DNA in discontinuities that had to be tackled with other methods. Paired-
end sequencing23 helped to link contigs into gapped scaffolds that could
3 Adaptor ligation
be followed up by directed sequencing to close gaps. Some problems were
4 Amplification only resolved by eye; scientists who were trained ‘finishers’ assessed the
quality and signed off on the assembled sequence of individual clones22.
Native DNA Although the process remained stable in its outlines, rapid-fire
improvements led to steady declines in the cost throughout the 1990s,
while parallel advances in computing increasingly replaced human
­decision making. By 2001, a small number of academic genome centres
were operating automated production lines generating up to 10 million
bases per day. Software for genome assembly matured both inside and
outside of the HGP, with tools, such as phrap, the TIGR assembler and the
5 Detection Cycle 1 Cycle 2 Cycle 3 Cycle 4
& 7 * $
Celera assembler, able to handle genomes of increasing c­ omplexity22,24,25.
A yearly doubling in capacity enabled the successful completion of
high-quality genomes beginning with Haemophilus influenza (around
3′... * $ & 7 $ * $ 7 & & * $ * & * 7 * $ ...5′ 2 Mb, 1995) followed quickly by Saccharomyces cerevisiae (around 12 Mb,
5′... & 7 * $ ... 1996) and Caenorhabditis elegans (around 100 Mb, 1998)26–28. The
HGP’s human genome, which is 30 times the size of C. elegans and with
Third generation sequencing
(Real-time, single molecule) &7*$7&7$**&7&*&$&7 much more repetitive content, came first as a draft (2001) and then as a
­finished sequence (2004)29,30. The HGP was paralleled by a private effort
to sequence a human genome by Craig Venter and Celera (2001)31 with
a whole-genome shotgun strategy piloted on Drosophila melanogaster
(around 175 Mb; 2000)32. The strategic contrasts between these projects
are further discussed below.
Figure 1 | DNA sequencing technologies. Schematic examples of first,
second and third generation sequencing are shown. Second generation
By 2004, instruments were churning out 600–700 bp at a cost of US$1
sequencing is also referred to as next-generation sequencing (NGS) per read, but creating additional improvements was an increasingly mar-
in the text. ginal exercise. Furthermore, with the completion of the HGP, the future
of large-scale DNA sequencing was unclear.
Scaling to the human genome
For the ‘hierarchical shotgun’ strategy that emerged as the workhorse Massively parallel DNA sequencing
of the Human Genome Project (HGP), large fragments of the human Throughout the 1980s and 1990s, several groups explored alternatives to
genome were cloned into bacterial artificial chromosomes (BACs). DNA electrophoretic sequencing. Although these efforts did not pay off until
from each BAC was fragmented, size-selected and sub-cloned. Individual after the HGP, within a decade of its completion, ‘massively parallel’ or
clones were picked and grown, and the DNA was isolated. The purified ‘next-generation’ DNA sequencing (NGS) almost completely superseded
DNA was used as a template for automated Sanger sequencing, the signal Sanger sequencing. NGS technologies sharply depart from electropho-
was extracted from laser-scanned images of the gels, and bases were called retic sequencing in several ways, but the key change is multiplexing.
to finally produce the sequence. The fact that this process involved many Instead of one tube per reaction, a complex library of DNA templates is
independent steps, each of which had to work well, led sceptics to doubt densely immobilized onto a two-dimensional surface, with all t­ emplates
it could ever be made efficient enough to sequence the human genome accessible to a single reagent volume. Rather than bacterial cloning,
at any reasonable cost. in vitro amplification generates copies of each template to be sequenced.
Indeed, as efforts to sequence larger genomes took shape, it became clear Finally, instead of measuring fragment lengths, sequencing comprises
that the scale and efficiency of each step needed to be vastly increased. This cycles of biochemistry (for example, polymerase-mediated incorpora-
was achieved in fits and spurts in the 1990s. Noteworthy improvements tion of ­fluorescently labelled nucleotides) and imaging, also known as
included: (1) a switch from dye-labelled primers to dye-­labelled termi- ‘sequencing-by-synthesis’ (SBS).
nators, which allowed one rather than four sequencing reactions17; (2) a Although amplification is not strictly necessary (for example,
mutant T7 DNA polymerase that more readily incorporated dye-labelled ­single-molecule SBS33–35), the dense multiplexing of NGS, with millions
terminators18; (3) linear amplification reactions, which greatly reduced to billions of immobilized templates, was largely enabled by clonal in vitro
template requirements and facilitated miniaturization19; (4) a magnetic amplification. The simplest approach, termed ‘polonies’ or ‘bridge ampli-
bead-based DNA purification method that simplified automation of prese- fication’, involves amplifying a complex template library with p ­ rimers
quencing steps20; (5) methods enabling sequencing of ­double-stranded immobilized on a surface, such that copies of each template remain tightly
DNA, which enabled the use of plasmid clones and therefore paired-end clustered36–39. Alternatively, clonal PCR can be performed in an ­emulsion,

3 4 6 | N A T U RE | V O L 5 5 0 | 1 9 o c t o b e r 2 0 1 7
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
 Review RESEARCH

De novo genome assembly

Sequence assembly of Mycoplasma genitalium by Margulies, Rothberg and 454
Short reads
overlaps
(ref. 40), and resequencing of phiX174 and a human BAC by Solexa54.
These studies demonstrated how useful even very short reads are, given
a ­reference genome to which to map them. Within three years, human
genome resequencing would become practical on the Solexa platform
Contigs with 35-bp paired reads55.
Long
In 2005, 454 released the first commercial NGS instrument. In the
range link wake of the HGP, large-scale sequencing was still the provenance of a few
genome centres. With 454 and other competing instruments that followed
closely after, individual laboratories could instantly access the capacity of
an entire HGP-era genome centre. This ‘democratization’ of sequencing
Scaffold
capacity had a profound impact on the culture and composition of the
genomics field, with new methods, results, genomes and other innova-
tions arising from all corners.
In contrast to the monopoly of Applied Biosystems during the HGP,
Chromosome
several companies, including 454 (acquired by Roche), Solexa (acquired
Genome resequencing Clinical applications (NIPT) by Illumina), Agencourt 47,48 (acquired by Applied Biosystems),
Individual Maternal
Helicos34,35 (founded by Quake), Complete Genomics43 (founded by
1 *$&7$*$7&&*$*&*7*$ blood Drmanac) and Ion Torrent46 (founded by Rothberg), intensely competed
plasma
on NGS, resulting in a rapidly changing landscape with new instruments
2 *$&7$*$7$&*$*&*7*$
Maternal that were flashily introduced at the annual Advances in Genome Biology
3 *$&*$*$7&&*&*&*7*$
DNA and Technology (AGBT) meeting in Marco Island, Florida. Between 2007
and 2012, the raw, per-base cost of DNA sequencing plummeted by four
...

7.5 *$&7$*$7&&*$*&*&*$ orders of magnitude56.

billion
Since 2012, the pace of improvement has slowed, as has the c­ ompetition.
Sites of
variation
*$&7$*$7&&*$*&*7*$ The 454, SOLiD and Helicos platforms are no longer being developed,
and the Illumina platform is dominant (although Complete Genomics43
remains a potential competitor). Nonetheless, it is astonishing to consider
Fetal DNA how far we have come since the inception of NGS in 2005. Read lengths,
Sequencers as counting devices
although still shorter than Sanger sequencing, are in the low hundreds
Ribosomal translation
of bases, and mostly over 99.9% accurate. Over a billion independent
reads, totalling a terabase of sequence, can be generated in two days by
one graduate student on one instrument (Illumina NovaSeq) for a few
RNA transcription
thousand dollars. This exceeds the approximately 23 gigabases that were
Chromatin accessibility generated for the HGP’s draft human genome by a factor of 40.

Figure 2 | DNA sequencing applications. Major categories of the Real-time, single-molecule sequencing
application of DNA sequencing include de novo genome assembly, Nearly all of the aforementioned platforms require template a­ mplification.
individual genome resequencing, clinical applications such as non-invasive However, the downsides of amplification include copying errors,
prenatal testing, and using sequencers as counting devices for a broad sequence-dependent biases and information loss (for example,
range of biochemical or molecular phenomena. methylation), not to mention added time and complexity. In an ideal
world, sequencing would be native, accurate and without read-length
such that copies of each template are immobilized on beads that are then limitations. To reach this goal, stretching back to the 1980s, a handful
arrayed on a surface for sequencing40–42. A third approach involves ­rolling of groups explored even more radical approaches than NGS. Many of
circle amplification in solution to generate clonal ­‘nanoballs’ that are these were dead ends, but at least two approaches were not, as these have
arrayed and sequenced43. recently given rise to real-time, single-molecule sequencing platforms
For SBS, there were three main strategies. The pyrosequencing that threaten to upend the field once again.
approach of Ronaghi and Nyrèn involves discrete, step-wise ­addition A first approach, initiated by Webb and Craighead and further
of each deoxynucleotide (dNTP). Incorporation of dNTPs releases ­developed by Korlach, Turner and Pacific Biosciences (PacBio), is to opti-
pyrophos­phate, which powers the generation of light by firefly cally observe polymerase-mediated synthesis in real time57,58. A zero mode
­luciferase44. With an analogous approach, natural dNTP incorporations waveguide, a hole less than half the wavelength of light, limits fluores-
can be detected with an ion-sensitive field effect transistor45,46. A second cent excitation to a tiny volume within which a single polymerase and its
strategy uses the specificity of DNA ligases to attach fluorescent oligonu- ­template reside. Therefore, only fluorescently labelled nucleotides incor-
cleotides to templates in a sequence-dependent manner41,43,47,48. A third porated into the growing DNA chain emit signals of sufficient duration
approach, which has proven the most durable, involves the stepwise, to be ‘called’. The engineered polymerase is highly processive; reads over
polymerase-mediated incorporation of fluorescently labelled deoxynu- 10 kb are typical, with some reads approaching 100 kb. The throughput of
cleotides33,34,49. Critical to the success of polymerase-mediated SBS, was PacBio is still over an order of magnitude less than the highest-throughput
the development of reversibly terminating, reversibly fluorescent dNTPs, NGS platforms, such as Illumina, but not so far from where NGS platforms
and a suitably engineered polymerase50, such that each template incor- were a few years ago. Error rates are very high (around 10%) but randomly
porates one and only one dNTP on each cycle. After imaging to deter- distributed. PacBio’s combination of minimal bias (for example, tolerance
mine which of four colours was incorporated by each template on the of extreme GC content), random errors, long reads and redundant cover-
surface, both blocking and fluorescent groups are removed to set up the age can result in de novo assemblies of unparalleled quality with respect
next extension51–53; this general approach was used by Solexa, founded to accuracy and contiguity, for many species exceeding what would be
by Balasubramanian and Klenerman in 1998. possible even with efforts similar to the HGP.
The first integrated NGS platforms came in 2005, with resequencing A second approach is nanopore sequencing. This concept, which was
of Escherichia coli by Shendure, Porreca, Mitra and Church41, de novo first hypothesized in the 1980s59–61, is based on the idea that patterns in

1 9 o c t o b e r 2 0 1 7 | V O L 5 5 0 | N A T U RE | 3 4 7
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
 RESEARCH Review

the commercial market is no longer monolithic and other technologies

Box 1 may successfully occupy important niches (for example, PacBio for
de novo assembly and ONT for portable sequencing). Neither NGS nor
The milestones listed below correspond to key developments in single-molecule methods have fully plateaued in cost and throughput,
the evolution of sequencing technologies. This is a large topic, and and there are additional concepts that are still in development, which are
we apologize for any omissions. not discussed here (for example, solid-state pores and electron micro-
Technical milestones scopy)70,71. Not all will work out, but as the above examples make clear,
1953: Sequencing of insulin protein2 transformative sequencing technologies can take decades to mature.
1965: Sequencing of alanine tRNA4
1968: Sequencing of cohesive ends of phage lambda DNA6 Applications of DNA sequencing
1977: Maxam–Gilbert sequencing9 The range and scope of DNA sequencing applications has also expanded
1977: Sanger sequencing8 over the past few decades, shaped in part by the evolving constraints
1981: Messing’s M13 phage vector12 of sequencing technologies. Below we review key areas of application
1986–1987: Fluorescent detection in electrophoretic ­including de novo genome assembly, individual genome ­resequencing,
sequencing14,15,17 sequencing in the clinic and the transformation of sequencers into
1987: Sequenase18 ­molecular counting devices. Some key milestones for the generation of
1988: Early example of sequencing by stepwise dNTP reference genomes and development of applications and software are
incorporation139 summarized in Box 2.
1990: Paired-end sequencing23
1992: Bodipy dyes140 De novo genome assembly
1993: In vitro RNA colonies37 For its first 25 years, the primary purpose of DNA sequencing was the
1996: Pyrosequencing44 partial or complete sequencing of genomes. Indeed, the inception of Sanger
1999: In vitro DNA colonies in gels38 sequencing in 1977 included the first genome (phiX174; 5.4 kb), essen-
2000: Massively parallel signature sequencing by ligation47 tially assembled by hand72. However, DNA sequencing was only one of
2003: Emulsion PCR to generate in vitro DNA colonies on beads42 several technologies that enabled assembly of larger genomes. If the DNA
2003: Single-molecule massively parallel sequencing-by-synthesis33,34 sequence was random, arbitrarily large genomes could be assembled to
2003: Zero-mode waveguides for single-molecule analysis57 completion solely based on fragment overlaps. However, it is not random,
2003: Sequencing by synthesis of in vitro DNA colonies in gels49 and the combination of repetitive sequences and technical biases makes
2005: Four-colour reversible terminators51–53 it impossible to obtain high-quality assemblies of large genomes from
2005: Sequencing by ligation of in vitro DNA colonies on beads41 kilobase-scale reads alone. Additional ‘contiguity information’ is required.
2007: Large-scale targeted sequence capture93–96 For the HGP29,30, these additional sources of contiguity information
2010: Direct detection of DNA methylation during single-molecule included the following. (1) Genetic maps, which were based on the seg-
sequencing65 regation of genetic polymorphisms through pedigrees, that provided
2010: Single-base resolution electron tunnelling through a solid- orthogonal information about the order of sequences locally and at the
state detector141 scale of chromosomes. (2) Physical maps, for which BACs were cloned,
2011: Semiconductor sequencing by proton detection142 restriction-enzyme ‘fingerprinted’ to identify overlaps and ordered into a
2012: Reduction to practice of nanopore sequencing143,144 ‘tiling path’ that spanned the genome. Clones were individually shotgun
2012: Single-stranded library preparation method for ancient DNA145 sequenced and assembled, thereby isolating different repeat copies from one
another, and then further ordered and assembled. (3) Paired-end sequenc-
ing, introduced by Ansorge in 199023, comprises sequencing into both ends
the flow of ions, which occur when a single-stranded DNA molecule of a DNA fragment of approximately known length, e­ ffectively linking those
passes through a narrow channel, will reveal the primary sequence of end-sequences. Depending on the cloning method, the spanned length
the strand. Decades of work were required to go from concept to reality. could range from a few kilobases to a few hundred kilobases. Sequence
Firstly, electric field-driven transport of DNA through a nanometre-scale coverage at 8–10-fold redundancy, coupled to these sources of contiguity
pore is so fast that the number of ions per nucleotide is insufficient to information, enabled not only genome assembly, but also improved quality
yield an adequate signal. Solutions have eventually been developed to to about 1 error per 100,000 bases for most of the genome. Additional,
these and other challenges, including interposing an enzyme as a ‘ratchet’, focused experiments were performed to fill the gaps or clarify ambiguities.
identifying and engineering improved nucleopore proteins, and better The Celera effort went straight to paired-end sequencing, ­eschewing
analytics of the resulting signals62. These advances recently culminated in physical maps as an intermediate31. An important advance was the
successful nanopore sequencing, in both academia63 and industry, most ­transition from greedy algorithms, such as phrap and the TIGR ­assembler,
prominently by Oxford Nanopore Technologies (ONT), founded by to the Celera assembler’s graph-based approach (overlap–layout–­
Bayley in 2005. Sequence read lengths of ONT are on par with or exceed consensus)22,24,25. Although Celera had a reasonable strategy for a draft
the reads generated by PacBio; with the longest obtained reads presently genome, because of the pervasiveness of repetitive sequences, it did not,
at 900 kilobases (ref. 64). A major differentiator from other sequencing by itself, result in a high-quality reference, such as the one produced by
technologies is the extreme portability of nanopore devices, which can be the HGP’s clone-based approach. The current human reference genome
as small as a memory (USB) stick, because they rely on the detection of descends from the HGP’s 2004 product30, with continuous work by the
electronic, rather than optical, signals. Important challenges remain (for Genome Reference Consortium to further improve it, including regular
example, errors may not be randomly distributed), but progress is rapid. releases of reference genome updates73.
Nucleic-acid sequencing would ideally also capture DNA With the advent of NGS in 2005, the number of de novo assemblies
­modifications. Indeed, both PacBio and nanopore sequencing have increased vastly. The seemingly disastrous combination of short reads
demonstrated the detection of native covalent modifications, such as and repetitive genomes was overcome by new assembly algorithms based
methylation64,65. Single-molecule methods also open up the intriguing on de Bruijn graphs (for example, EULER and Velvet)74,75. Nonetheless,
possibility of directly sequencing RNA66,67 or even proteins68–71. particularly when applied to larger genomes and when compared to the
Since 1977, DNA-sequencing technology has evolved at a fast pace genomes of the HGP, their quality was, on average, quite poor. Although
and the landscape continues to change shift under our feet. Although shorter read lengths are partly to blame, this is usually overstated. Instead,
Illumina is presently the dominant supplier of sequencing instruments, a principal reason for the poorer quality was the paucity of contiguity

3 4 8 | N A T U RE | V O L 5 5 0 | 1 9 o c t o b e r 2 0 1 7
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
 Review RESEARCH

Box 2
The milestones listed below correspond to key developments in 2005: Galaxy161
the availability of new reference genomes, new sequencing-related 2007: NCBI Short Read Archive
computational tools and the applications of DNA sequencing in new 2008: ALLPATHS162
ways or to new areas. These are large topics, and we apologize for 2008: Velvet75
any omissions. 2009: Bowtie83
Genome milestones 2009: BWA82
1977: Bacteriophage ΦX174 (ref. 72) 2009: SAMtools84
1982: Bacteriophage lambda13 2009: BreakDancer163
1995: Haemophilus influenzae26 2009: Pindel164
1996: Saccharomyces cerevisiae27 2009: TopHat115
1998: Caenorhabditis elegans28 2010: SOAPdenovo165
2000: Drosophila melanogaster32 2010: GATK85
2000: Arabidopsis thaliana146 2010: Cufflinks116
2001: Homo sapiens29–31 2011: Integrated Genomics Viewer166
2002: Mus musculus147 2013: HGAP/Quiver167
2004: Rattus norvegicus148 2017: Canu81
2005: Pan troglodytes149 Application milestones
2005: Oryza sativa150 1977: Genome sequencing72
2007: Cyanidioschyzon merolae126 1982: Shotgun sequencing13
2009: Zea mays151 1983, 1991: Expressed sequence tags107,108
2010: Neanderthal88 1995: Serial analysis of gene expression109
2012: Denisovan145 1998: Large-scale human SNP discovery168
2013: The HeLa cell line152,153 2004: Metagenome assembly122
2013: Danio rerio154 2005: Bacterial genome resequencing with NGS40,41
2017: Xenopus laevis155 2007: Chromatin immunoprecipitation followed by sequencing
Computational milestones (ChIP–seq) using NGS117
1981: Smith–Waterman156 2007–2008: Human genome and cancer genome resequencing using
1982: GenBank (https://www.ncbi.nlm.nih.gov/genbank/statistics/) NGS55,90–92
1990: BLAST16 2008: RNA-seq using NGS110–114
1995: TIGR assembler24 2008: Chromatin accessibility using NGS118
1996: RepeatMasker 2009: Exome resequencing using NGS97
1997: GENSCAN157 2009: Ribosome profiling using NGS119
1998: phred, phrap, consed22 2010: Completion of Phase I of the 1000 Genomes Project98
2000: Celera assembler25 2010: De novo assembly of a large genome from short reads169
2001: Bioconductor 2011: Haplotype-resolved human genome resequencing using
2001: EULER74 NGS170,171
2002: BLAT158 2016: Human genome de novo assembly with PacBio172
2002: UCSC Genome Browser159 2017: Human genome de novo assembly with nanopore64
2002: Ensembl160

methods to complement NGS. Paired-end sequencing was possible with Genome resequencing
NGS, but in vitro library methods were more restricted with respect to the After the HGP, a clear next step was to catalogue genetic variation among
distances that could be spanned. Furthermore, the field lacked ‘massively humans, that is, ‘resequencing’. Because Sanger sequencing costs remained
parallel’ equivalents of genetic and physical maps. high, resequencing was primarily used to discover common variants,
This ‘dark’ period notwithstanding, there are good reasons to be opti- which were then cost-effectively genotyped with microarrays to facilitate
mistic about the future of de novo assembly. Firstly, in vitro methods that genome-wide association studies. The rallying cry for changing this was
subsample high molecular weight (HMW) genomic fragments, analogous the ‘US$1,000 human genome’, the ambitious goal of the resequencing
to hierarchical shotgun sequencing, have recently been developed76,77. of individuals at a cost nearly one-million-fold below that of assembling
Secondly, methods, such as Hi-C (genome-wide chromosome conforma- the first human genome. The US$1,000 genome concept was discussed as
tion capture) and optical mapping, provide scalable, cost-effective means early as 2001 (at the University of California, Santa Cruz Human Genome
of scaffolding genomes into chromosome-scale assemblies78–80. Finally, Symposium (http://genomesymposium.ucsc.edu)), when NGS strategies
the read lengths of PacBio and ONT sequencing have risen to hundreds barely existed, and was formalized a few years later by the Revolutionary
of kilobases, and are now more limited by the preparation of HMW DNA Sequencing Technologies program of the National Human Genome
DNA than by the sequencing itself. The absence of cloning or amplifi- Research Institute (NHGRI). The commitment of US$220 million in
cation steps in single-molecule sequencing pays off, as shown by high-­ funding to over 40 academic and 27 commercial entities has helped to
quality PacBio de novo assemblies of bacterial genomes with extreme GC drive much of the technological development described above, including
content. Long reads have resulted in a re-emergence of strategies used direct or indirect support of nearly every successful commercial platform.
by the Celera assembler, improved to deal with the high error rates Resequencing, that is, mapping sequence reads to a reference genome
and m ­ ultiple platforms81. By combining long reads and even longer- to identify genetic variants, is a very different task than genome assembly.
range contiguity information (for example, subsampling HMW DNA, New algorithms, such as Bowtie and Burrows–Wheeler Aligner (BWA),
­chromatin proximity, optical maps and so on), de novo genome assemblies borrowed concepts from data-compression techniques to enable millions
of the quality of the original human reference genome using ‘post-Sanger’ of reads to be efficiently mapped to the reference genome82,83. Redundant
approaches are finally within sight73,80. coverage (for example, 30-fold) is necessary to identify heterozygous

1 9 o c t o b e r 2 0 1 7 | V O L 5 5 0 | N A T U RE | 3 4 9
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
 RESEARCH Review

variants as well as to distinguish sequencing errors from bona fide Our understanding of cancer, fundamentally a disease of the
­variants. Popular packages, initially SAMtools and later GATK, adapted genome, is gradually being transformed by DNA sequencing. Large-
the confidence framework of phred to NGS bases, reads and variants84,85. scale ­resequencing has laid bare the extraordinary genetic hetero-­
Short reads, particularly when paired, can be uniquely mapped to most geneity of cancers, ­effectively defining a molecular taxonomy106. DNA
of the human genome. But most is not all, and a problem of short-read sequencing is i­ mpacting clinical cancer care by: (1) suggesting targeted
resequencing is that variants in repetitive regions and structural ­variants therapies, based on the mutations present in an individual cancer;
are routinely missed. The extent of this shortcoming is quantified by (2) enabling non-­invasive diagnosis or monitoring by sequencing of
recent studies that resequence human genomes with PacBio86. A second tumour-­released circulating cells or cell-free DNA; (3) identifying
aspect of incompleteness relates to phase relationships between v­ ariants ­cancer-specific, ­protein-altering mutations that may serve as neoanti-
in a diploid genome, that is, haplotypes87. Fortunately, haplotypes are gens for ‘personal vaccines’. Although, the success stories in each of these
recovered by many of the same methods that enable contiguity for areas are still few and far between, relative to the overall burden of cancer,
de novo NGS assemblies (and ideally, even de novo assemblies would be progress is clearly being made.
haplotype-resolved)77. Although still not broadly used, these methods are
becoming increasingly scalable. Sequencers as a molecule counting device
The HGP’s human genome was constructed from a mosaic of DNA While ‘expressed sequenced tags’107 were considered a shortcut to
donors, but mostly derives from one individual, from Buffalo, New gene discovery as early as 1983108, it was SAGE (serial analysis of gene
York, who had roughly equal parts European and African ancestry88. ­expression; 1995) that introduced the idea of sequencing to ‘digitally
The first individual to have their whole genome resequenced was quantify’ gene expression109. SAGE concatenated cDNA-derived tags for
Craig Venter in 2007, one of the subjects of the Celera genome, which Sanger sequencing, with tags that are just long enough to map to a gene.
was supplemented with additional data89. This was quickly followed As early as 2000, Brenner and Lynx Therapeutics demonstrated ‘massively
in 2008 by the genome of Jim Watson on 454 (ref. 90), and then the parallel signature sequencing’ of cDNA tags, an important forerunner of
genomes of two ­anonymous individuals55,91 and the germline and NGS47. However, this concept was not widely adopted until the develop-
tumour genome of a patient92 on Solexa/Illumina, and five individuals ment of RNA sequencing (RNA-seq) by five groups in 2008. RNA-seq
on Complete Genomics43. In this period, whole-genome sequencing uses NGS to quantify and characterize the transcriptome by shotgun
(WGS) remained too expensive for most groups to scale, motivating sequencing of either full-length or 3′​ends of cDNA110–114. RNA-seq has
the development of targeted capture methods93–96 and then whole-­ marked advantages over microarrays, the most notable of which is that
exome sequencing (WES), that is, selective sequencing of the 1–2% of transcript counts lead to straightforward statistics relative to analogue,
the genome that is encodes proteins97. hybridization-based signals, facilitated by new software packages, such
As costs approached US$1,000 for WGS56 and a few hundred d ­ ollars as TopHat and Cufflinks115,116.
for WES, the pace at which individual humans are resequenced has Also around 2008, small laboratories that were early adopters of
accelerated. The 1000 Genomes Project, launched in 2008, released NGS developed ‘digital quantification’ methods for transcription-factor
low-coverage WGS of a few hundred ­individuals in 2010 and a few ­binding117, chromatin accessibility118 and translation119. In the following
thousand individuals in 201598,99. The Exome Sequencing Project decade, hundreds of protocols were developed that facilitate the use of
released over 6,500 exomes in 2013100. The recently released Genome DNA sequencing as a ‘molecule counter’ for the characterization of a
Aggregation Database (http://gnomad.broadinstitute.org/) includes remarkable range of biochemical or molecular phenomena, ­including
more than 120,000 exomes and over 15,000 genomes. The Genomics transcription, translation, DNA replication, the secondary structure of
England (https://www.genomicsengland.co.uk/), GenomeAsia100K RNA, chromosome conformation, nucleic-acid modifications, post-­
(http://www.genomeasia100k.com/) and NHLBI TOPMed (Trans- translational modifications, nucleic acid–protein interactions and
Omics for Precision Medicine, https://www.nhlbiwgs.org/) ­projects ­protein–protein interactions. These are catalogued in other reviews and
each aim to complete WGS on more than 100,000 individuals resources (ref. 120 and http://enseqlopedia.com/).
within the next year or two. Given that these projects represent a fraction The use of sequencers as molecule-counting devices was immediately
of all sequencing being conducted, it is plausible that the genomes of over immensely popular, and probably had a larger role than assembly or
one million humans have already been resequenced by WES or WGS. resequencing in driving the widespread adoption of NGS in biomedical
research. DNA sequencers are increasingly to the molecular biologist
Clinical applications of sequencing what a microscope is to the cellular biologist—a basic and essential tool
Our ability to sequence human genomes has vastly outpaced our for making measurements. In the long run, this may prove to be greatest
­ability to interpret genetic variation, which partly explains why clinical impact of DNA sequencing.
­medicine has yet to wholeheartedly embrace WGS. Nonetheless, there are
some areas in which DNA sequencing is already proving clinically useful, Metagenome sequencing
three of which we highlight here. Shotgun sequencing of complex communities of microorganisms121–123,
The most unexpected area of the clinical impact of DNA sequencing for example, metagenome sequencing of environmental or human
has been non-invasive prenatal testing (NIPT, see Fig. 2). Pioneering microbiomes, has emerged as a field of its own, bringing with it unique
­studies by Lo and Quake in 2008 have demonstrated that the simple challenges with respect to assembly, resequencing and counting. Other
counting of DNA fragments released into the maternal circulation by reviews have recently covered this topic124,125.
a fetus during pregnancy can detect chromosomal aneuploidies101,102.
Screening tests that were based on this strategy were adopted faster than The future of DNA sequencing
any molecular test in history, and several million pregnant women around In the long view of scientific history, DNA sequencing remains a young
the world have already benefited from low-pass WGS for NIPT. technology. Here, we briefly consider its future in a few existing or
An early application of WES was to rapidly discover new genes for, ­emerging areas.
and to diagnose patients affected by, Mendelian ­disorders97,103. This was
quickly followed by the discovery that ­substantial proportions of neu- Genome diversity
rodevelopmental disorders are attributable to de novo mutations in coding A 100% complete genome, that is, the telomere-to-telomere sequence
sequences104. WES is increasingly used as a p ­ rimary tool for diagnosing for each chromosome with no gaps or ambiguities, has been achieved
Mendelian and neurodevelopmental d ­ isorders, ­particularly in paedi- for ­possibly only one eukaryote so far126. As sequencing technologies
atric populations, with the rate of diagnosis of patients with suspected continue to evolve, we are optimistic that we will resolve challenging
Mendelian disease by WES at 25% and rising105. regions of additional genomes (for example, centromeres). There are

3 5 0 | N A T U RE | V O L 5 5 0 | 1 9 o c t o b e r 2 0 1 7
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
 Review RESEARCH

­ illions of living species on earth (and far more extinct species), each
m 5. Sanger, F., Brownlee, G. G. & Barrell, B. G. A two-dimensional fractionation
with a genome waiting to be sequenced, as well as countless micro­ procedure for radioactive nucleotides. J. Mol. Biol. 13, 373–398 (1965).
6. Wu, R. & Kaiser, A. D. Structure and base sequence in the cohesive ends of
biomes and m ­ etagenomes. A comprehensive view of genomic diversity bacteriophage lambda DNA. J. Mol. Biol. 35, 523–537 (1968).
may prove useful in surprising ways, for example, for protein structure 7. Gilbert, W. & Maxam, A. The nucleotide sequence of the lac operator. Proc. Natl
­determination127. Acad. Sci. USA 70, 3581–3584 (1973).
8. Sanger, F., Nicklen, S. & Coulson, A. R. DNA sequencing with chain-terminating
inhibitors. Proc. Natl Acad. Sci. USA 74, 5463–5467 (1977).
Population-scale resequencing Refs 8, 9: The seminal papers by Sanger, Nicklen & Coulson and Maxam &
We are approaching the milestone where approximately 0.1% of ­living Gilbert describing the first widely adopted methods for DNA sequencing.
9. Maxam, A. M. & Gilbert, W. A new method for sequencing DNA. Proc. Natl Acad.
humans will have had their genomes resequenced to some degree, while Sci. USA 74, 560–564 (1977).
resequencing of the genomes of our ancestors and other ­hominins 10. Maniatis, T., Jeffrey, A. & van deSande, H. Chain length determination of small
is reshaping our understanding of human history88. The number of double- and single-stranded DNA molecules by polyacrylamide gel
electrophoresis. Biochemistry 14, 3787–3794 (1975).
de novo point mutations occurring in recent generations vastly exceeds 11. Staden, R. A strategy of DNA sequencing employing computer programs.
the number of nucleotides in the human genome. Eventually, aggregating Nucleic Acids Res. 6, 2601–2610 (1979).
tens of millions of genomes may enable a nucleotide-level footprint of the 12. Messing, J., Crea, R. & Seeburg, P. H. A system for shotgun DNA sequencing.
human genome (that is, observing all heterozygous variants ­compatible Nucleic Acids Res. 9, 309–321 (1981).
13. Sanger, F., Coulson, A. R., Hong, G. F., Hill, D. F. & Petersen, G. B. Nucleotide
with life). DNA sequencing also is increasingly useful for forensics, sequence of bacteriophage lambda DNA. J. Mol. Biol. 162, 729–773 (1982).
­without necessarily requiring a sample from the identified individual128. 14. Smith, L. M. et al. Fluorescence detection in automated DNA sequence
analysis. Nature 321, 674–679 (1986).
15. Connell, C. et al. Automated DNA sequence analysis. Biotechniques 5,
Developmental biology 342–348 (1987).
We each develop from a single cell into a highly organized mass of 16. Altschul, S. F., Gish, W., Miller, W., Myers, E. W. & Lipman, D. J. Basic local
­trillions of cells. However, our understanding of development remains alignment search tool. J. Mol. Biol. 215, 403–410 (1990).
17. Prober, J. M. et al. A system for rapid DNA sequencing with fluorescent
coarse. Recent technologies enable scalable, sequencing-based chain-terminating dideoxynucleotides. Science 238, 336–341 (1987).
­profiling of single cells. Although popular approaches are ex vivo (for 18. Tabor, S. & Richardson, C. C. DNA sequence analysis with a modified
example, single-cell RNA-seq), a more radical approach is to p ­ erform bacteriophage T7 DNA polymerase. Proc. Natl Acad. Sci. USA 84, 4767–4771
(1987).
RNA or protein ­s equencing in situ, thereby retaining the spatial 19. Craxton, M. Linear amplification sequencing, a powerful method for
­context129,130. Other emerging strategies use in vivo genome editing to sequencing DNA. Methods 3, 20–26 (1991).
track cell-­lineage ­relationships131 or transport barcodes to catalogue 20. DeAngelis, M. M., Wang, D. G. & Hawkins, T. L. Solid-phase reversible
neuronal ­connections132. Editing of DNA can potentially be used to immobilization for the isolation of PCR products. Nucleic Acids Res. 23,
4742–4743 (1995).
record b ­ iological events more generally, for example, to monitor gene 21. Zhang, J. et al. Use of non-cross-linked polyacrylamide for four-color DNA
­expression133 or calcium134. sequencing by capillary electrophoresis separation of fragments up to 640
bases in length in two hours. Anal. Chem. 67, 4589–4593 (1995).
22. Green, P. phred, phrap, consed. http://www.phrap.org/phredphrapconsed.
Real-time, portable sensors html (2017).
Nanopore sequencers currently have a mass of 70 g and yield data within phred introduced quantitative, reliable metrics for base quality, substituting
30 min of sample application. One can imagine disseminated networks human judgement with computers, a process that occurred repeatedly over
the course of the HGP.
of nanopore sequencers enabling ‘universal monitoring’ of nucleic acids, 23. Edwards, A. et al. Automated DNA sequencing of the human HPRT locus.
in environmental settings and in everyday human life, for example, Genomics 6, 593–608 (1990).
fine-grained tracking of our air, food and body, potentially streaming 24. Sutton, G. G., White, O., Adams, M. D. & Kerlavage, A. R. TIGR assembler: a new
tool for assembling large shotgun sequencing projects. Genome Sci. Technol.
data from millions of devices and integrating with GPS and audio- 1, 9–19 (1995).
visual data. 25. Myers, E. W. et al. A whole-genome assembly of Drosophila. Science 287,
2196–2204 (2000).
The Celera assembler introduced an overlap–layout–consensus approach to
Unconventional uses deal with the problems posed by repeats and the millions of reads needed to
DNA-sequencing technologies will probably prove useful in additional, produce a reliable assembly.
surprising ways. For example, NGS has recently been used to recover 26. Fleischmann, R. D. et al. Whole-genome random sequencing and assembly of
large amounts of data encoded in synthetic DNA135. Nanopores may find Haemophilus influenzae Rd. Science 269, 496–512 (1995).
27. Goffeau, A. et al. Life with 6000 genes. Science 274, 546–567 (1996).
uses beyond sequencing, for example, for monitoring analyte binding136, 28. The C. elegans Sequencing Consortium. Genome sequence of the nematode
chemical nanomachines137 or protein folding/unfolding138. C. elegans: a platform for investigating biology. Science 282, 2012–2018
(1998).
29. International Human Genome Sequencing Consortium. Initial sequencing and
DNA sequencing as the new microscope analysis of the human genome. Nature 409, 860–921 (2001).
It has been about 400 years since the invention of light microscopy, a Refs 29–31: The HGP and Celera produced draft sequences of the human
technology which continues to be used and to evolve. By comparison, genome with the HGP later publishing a more complete, relatively error-free
reference.
it has been only 40 years since the invention of DNA sequencing; the 30. International Human Genome Sequencing Consortium. Finishing the
­technologies for which are likely to also continue to develop in the coming euchromatic sequence of the human genome. Nature 431, 931–945
decades and centuries. On the basis of how quickly it has transformed (2004).
biomedical research, and is beginning to transform clinical medicine, 31. Venter, J. C. et al. The sequence of the human genome. Science 291,
1304–1351 (2001).
we predict that DNA sequencing will have a longevity and impact on par 32. Adams, M. D. et al. The genome sequence of Drosophila melanogaster. Science
with or exceeding that of the microscope. 287, 2185–2195 (2000).
33. Balasubramanian, S., Klenerman, D. & Barnes, C. Arrayed polynucleotides and
received 13 July; accepted 21 September 2017. their use in genome analysis. Patent US20030022207 (2003).
Published online 11 October 2017. 34. Braslavsky, I., Hebert, B., Kartalov, E. & Quake, S. R. Sequence information can
be obtained from single DNA molecules. Proc. Natl Acad. Sci. USA 100,
3960–3964 (2003).
1. Sanger, F. Sequences, sequences, and sequences. Annu. Rev. Biochem. 57, 35. Harris, T. D. et al. Single-molecule DNA sequencing of a viral genome. Science
1–28 (1988). 320, 106–109 (2008).
2. Sanger, F. Nobel lecture: the chemistry of insulin. https://www.nobelprize.org/ 36. Adams, C. P. & Kron, S. J. Method for performing amplification of nucleic acid
nobel_prizes/chemistry/laureates/1958/sanger-lecture.html (2017). with two primers bound to a single solid support. Patent US5641658 (1997).
3. Edman, P. Method for determination of the amino acid sequence in peptides. 37. Chetverina, H. V. & Chetverin, A. B. Cloning of RNA molecules in vitro. Nucleic
Acta Chem. Scand. 4, 283–293 (1950). Acids Res. 21, 2349–2353 (1993).
4. Holley, R. W. et al. Structure of a ribonucleic acid. Science 147, 1462–1465 38. Mitra, R. D. & Church, G. M. In situ localized amplification and contact replication
(1965). of many individual DNA molecules. Nucleic Acids Res. 27, e34–e39 (1999).

1 9 o c t o b e r 2 0 1 7 | V O L 5 5 0 | N A T U RE | 3 5 1
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
 RESEARCH Review

39. Adessi, C. et al. Solid phase DNA amplification: characterisation of primer 70. Wilson, J., Sloman, L., He, Z. & Aksimentiev, A. Graphene nanopores for protein
attachment and amplification mechanisms. Nucleic Acids Res. 28, e87 (2000). sequencing. Adv. Funct. Mater. 26, 4830–4838 (2016).
40. Margulies, M. et al. Genome sequencing in microfabricated high-density 71. Di Ventra, M. & Taniguchi, M. Decoding DNA, RNA and peptides with quantum
picolitre reactors. Nature 437, 376–380 (2005). tunnelling. Nat. Nanotechnol. 11, 117–126 (2016).
Refs 40, 41: These papers described the first integrated systems for 72. Sanger, F. et al. Nucleotide sequence of bacteriophage phi X174 DNA. Nature
next-generation DNA sequencing. 265, 687–695 (1977).
41. Shendure, J. et al. Accurate multiplex polony sequencing of an evolved 73. Schneider, V. A. et al. Evaluation of GRCh38 and de novo haploid genome
bacterial genome. Science 309, 1728–1732 (2005). assemblies demonstrates the enduring quality of the reference assembly.
42. Dressman, D., Yan, H., Traverso, G., Kinzler, K. W. & Vogelstein, B. Transforming Genome Res. 27, 849–864 (2017).
single DNA molecules into fluorescent magnetic particles for detection 74. Pevzner, P. A., Tang, H. & Waterman, M. S. An Eulerian path approach to DNA
and enumeration of genetic variations. Proc. Natl Acad. Sci. USA 100, fragment assembly. Proc. Natl Acad. Sci. USA 98, 9748–9753 (2001).
8817–8822 (2003). 75. Zerbino, D. R. & Birney, E. Velvet: algorithms for de novo short read assembly
43. Drmanac, R. et al. Human genome sequencing using unchained base reads on using de Bruijn graphs. Genome Res. 18, 821–829 (2008).
self-assembling DNA nanoarrays. Science 327, 78–81 (2010). 76. Adey, A. et al. In vitro, long-range sequence information for de novo genome
44. Ronaghi, M., Karamohamed, S., Pettersson, B., Uhlén, M. & Nyrén, P. Real-time assembly via transposase contiguity. Genome Res. 24, 2041–2049 (2014).
DNA sequencing using detection of pyrophosphate release. Anal. Biochem. 77. Mostovoy, Y. et al. A hybrid approach for de novo human genome sequence
242, 84–89 (1996). assembly and phasing. Nat. Methods 13, 587–590 (2016).
45. Toumazou, C. & Purushothaman, S. Sensing apparatus and method. Patent 78. Burton, J. N. et al. Chromosome-scale scaffolding of de novo genome assemblies
US7686929 (2004). based on chromatin interactions. Nat. Biotechnol. 31, 1119–1125 (2013).
46. Rothberg, J. M., Hinz, W., Johnson, K. L. & Bustillo, J. Apparatus for measuring 79. Kaplan, N. & Dekker, J. High-throughput genome scaffolding from in vivo DNA
analytes using large scale FET arrays. Patent EP2639579 (2016). interaction frequency. Nat. Biotechnol. 31, 1143–1147 (2013).
47. Brenner, S. et al. Gene expression analysis by massively parallel signature 80. Bickhart, D. M. et al. Single-molecule sequencing and chromatin conformation
sequencing (MPSS) on microbead arrays. Nat. Biotechnol. 18, 630–634 capture enable de novo reference assembly of the domestic goat genome.
(2000). Nat. Genet. 49, 643–650 (2017).
48. McKernan, K. J. et al. Sequence and structural variation in a human genome 81. Koren, S. et al. Canu: scalable and accurate long-read assembly via adaptive
uncovered by short-read, massively parallel ligation sequencing using k-mer weighting and repeat separation. Genome Res. 27, 722–736 (2017).
two-base encoding. Genome Res. 19, 1527–1541 (2009). 82. Li, H. & Durbin, R. Fast and accurate short read alignment with Burrows–
49. Mitra, R. D., Shendure, J., Olejnik, J., Edyta-Krzymanska-Olejnik, & Wheeler transform. Bioinformatics 25, 1754–1760 (2009).
Church, G. M. Fluorescent in situ sequencing on polymerase colonies. 83. Langmead, B., Trapnell, C., Pop, M. & Salzberg, S. L. Ultrafast and memory-
Anal. Biochem. 320, 55–65 (2003). efficient alignment of short DNA sequences to the human genome. Genome
50. Ost, T. B. et al. Improved polymerases. Patent WO2006120433 (2006). Biol. 10, R25 (2009).
51. Ruparel, H. et al. Design and synthesis of a 3′​-O-allyl photocleavable 84. Li, H. et al. The sequence alignment/map format and SAMtools. Bioinformatics
fluorescent nucleotide as a reversible terminator for DNA sequencing by 25, 2078–2079 (2009).
synthesis. Proc. Natl Acad. Sci. USA 102, 5932–5937 (2005). 85. McKenna, A. et al. The genome analysis toolkit: a MapReduce framework for
52. Seo, T. S. et al. Four-color DNA sequencing by synthesis on a chip using analyzing next-generation DNA sequencing data. Genome Res. 20,
photocleavable fluorescent nucleotides. Proc. Natl Acad. Sci. USA 102, 1297–1303 (2010).
5926–5931 (2005). 86. Chaisson, M. J. et al. Resolving the complexity of the human genome using
53. Barnes, C., Balasubramanian, S., Liu, X., Swerdlow, H. & Milton, J. Labelled single-molecule sequencing. Nature 517, 608–611 (2015).
nucleotides. Patent US7057026 (2006). 87. Snyder, M. W., Adey, A., Kitzman, J. O. & Shendure, J. Haplotype-resolved
54. Smith, T. J. Cloned single molecule sequencing with reversible terminator genome sequencing: experimental methods and applications. Nat. Rev. Genet.
chemistry. Genome Sequencing and Analysis Conference (2015). 16, 344–358 (2015).
55. Bentley, D. R. et al. Accurate whole human genome sequencing using 88. Green, R. E. et al. A draft sequence of the Neandertal genome. Science 328,
reversible terminator chemistry. Nature 456, 53–59 (2008). 710–722 (2010).
Advances in sequencing-by-synthesis culminated in the Solexa, later 89. Levy, S. et al. The diploid genome sequence of an individual human. PLoS Biol.
Illumina, platform, which quickly became, and remains today, the most 5, e254 (2007).
widely used sequencing instrument. 90. Wheeler, D. A. et al. The complete genome of an individual by massively
56. Wetterstrand, K. DNA sequencing costs: data from the NHGRI Genome parallel DNA sequencing. Nature 452, 872–876 (2008).
Sequencing Program (GSP). http://www.genome.gov/sequencingcostsdata 91. Wang, J. et al. The diploid genome sequence of an Asian individual. Nature
(2017). 456, 60–65 (2008).
57. Levene, M. J. et al. Zero-mode waveguides for single-molecule analysis at high 92. Ley, T. J. et al. DNA sequencing of a cytogenetically normal acute myeloid
concentrations. Science 299, 682–686 (2003). leukaemia genome. Nature 456, 66–72 (2008).
One of earliest real time observations of DNA synthesis in single molecules, 93. Albert, T. J. et al. Direct selection of human genomic loci by microarray
using fluorescently labelled nucleotides and a DNA polymerase anchored in hybridization. Nat. Methods 4, 903–905 (2007).
zero-mode waveguides, which with further development led to the PacBio 94. Okou, D. T. et al. Microarray-based genomic selection for high-throughput
platform. resequencing. Nat. Methods 4, 907–909 (2007).
58. Eid, J. et al. Real-time DNA sequencing from single polymerase molecules. 95. Porreca, G. J. et al. Multiplex amplification of large sets of human exons.
Science 323, 133–138 (2009). Nat. Methods 4, 931–936 (2007).
59. Deamer, D., Akeson, M. & Branton, D. Three decades of nanopore sequencing. 96. Hodges, E. et al. Genome-wide in situ exon capture for selective resequencing.
Nat. Biotechnol. 34, 518–524 (2016). Nat. Genet. 39, 1522–1527 (2007).
60. Bayley, H. Nanopore sequencing: from imagination to reality. Clin. Chem. 61, 97. Ng, S. B. et al. Targeted capture and massively parallel sequencing of 12
25–31 (2015). human exomes. Nature 461, 272–276 (2009).
61. Church, G., Deamer, D. W., Branton, D., Baldarelli, R. & Kasianowicz, J. Refs 97, 103, 106: Targeting all coding sequences or the exome, by PCR and
Characterization of individual polymer molecules based on monomer- later by exome capture, facilitated the direct discovery of cancer driver
interface interactions. Patent US5795782 (1998). genes and Mendelian disease genes.
The concept of ssDNA modulating an electronic signal while moving through 98. The 1000 Genomes Project Consortium. A map of human genome variation
a membrane pore led eventually to practical nanopore sequencing. from population-scale sequencing. Nature 467, 1061–1073 (2010).
62. Branton, D. et al. The potential and challenges of nanopore sequencing. 99. The 1000 Genomes Project Consortium. A global reference for human genetic
Nat. Biotechnol. 26, 1146–1153 (2008). variation. Nature 526, 68–74 (2015).
63. Laszlo, A. H. et al. Decoding long nanopore sequencing reads of natural DNA. 100. Fu, W. et al. Analysis of 6,515 exomes reveals the recent origin of most human
Nat. Biotechnol. 32, 829–833 (2014). protein-coding variants. Nature 493, 216–220 (2013).
64. Jain, M. et al. Nanopore sequencing and assembly of a human genome with 101. Chiu, R. W. et al. Noninvasive prenatal diagnosis of fetal chromosomal
ultra-long reads. Preprint at https://www.biorxiv.org/content/ aneuploidy by massively parallel genomic sequencing of DNA in maternal
early/2017/04/20/128835 (2017). plasma. Proc. Natl Acad. Sci. USA 105, 20458–20463 (2008).
65. Flusberg, B. A. et al. Direct detection of DNA methylation during single- 102. Fan, H. C., Blumenfeld, Y. J., Chitkara, U., Hudgins, L. & Quake, S. R. Noninvasive
molecule, real-time sequencing. Nat. Methods 7, 461–465 (2010). diagnosis of fetal aneuploidy by shotgun sequencing DNA from maternal
66. Smith, A. M., Jain, M., Mulroney, L., Garalde, D. R. & Akeson, M. Reading blood. Proc. Natl Acad. Sci. USA 105, 16266–16271 (2008).
canonical and modified nucleotides in 16S ribosomal RNA using nanopore 103. Choi, M. et al. Genetic diagnosis by whole exome capture and massively parallel
direct RNA sequencing. Preprint at https://www.biorxiv.org/content/ DNA sequencing. Proc. Natl Acad. Sci. USA 106, 19096–19101 (2009).
early/2017/04/29/132274 (2017). 104. Vissers, L. E. L. M., Gilissen, C. & Veltman, J. A. Genetic studies in intellectual
67. Garalde, D. R. et al. Highly parallel direct RNA sequencing on an array of disability and related disorders. Nat. Rev. Genet. 17, 9–18 (2016).
nanopores. Preprint at https://www.biorxiv.org/content/early/2016/ 105. Yang, Y. et al. Molecular findings among patients referred for clinical
08/12/068809 (2016). whole-exome sequencing. J. Am. Med. Assoc. 312, 1870–1879 (2014).
68. Nivala, J., Marks, D. B. & Akeson, M. Unfoldase-mediated protein translocation 106. Wood, L. D. et al. The genomic landscapes of human breast and colorectal
through an α​-hemolysin nanopore. Nat. Biotechnol. 31, 247–250 (2013). cancers. Science 318, 1108–1113 (2007).
69. Zhao, Y. et al. Single-molecule spectroscopy of amino acids and peptides by 107. Adams, M. D. et al. Complementary DNA sequencing: expressed sequence
recognition tunnelling. Nat. Nanotechnol. 9, 466–473 (2014). tags and human genome project. Science 252, 1651–1656 (1991).

3 5 2 | N A T U RE | V O L 5 5 0 | 1 9 o c t o b e r 2 0 1 7
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
 Review RESEARCH

108. Putney, S. D., Herlihy, W. C. & Schimmel, P. A new troponin T and cDNA clones 143. Manrao, E. A. et al. Reading DNA at single-nucleotide resolution with a mutant
for 13 different muscle proteins, found by shotgun sequencing. Nature 302, MspA nanopore and phi29 DNA polymerase. Nat. Biotechnol. 30, 349–353
718–721 (1983). (2012).
109. Velculescu, V. E., Zhang, L., Vogelstein, B. & Kinzler, K. W. Serial analysis of gene 144. Cherf, G. M. et al. Automated forward and reverse ratcheting of DNA in a
expression. Science 270, 484–487 (1995). nanopore at 5-Å precision. Nat. Biotechnol. 30, 344–348 (2012).
The SAGE method captures 3′ tags from mRNAs, therefore introducing the 145. Meyer, M. et al. A high-coverage genome sequence from an archaic Denisovan
idea of using a DNA sequencer to count molecules, an idea that has individual. Science 338, 222–226 (2012).
exploded with the later introduction of RNA-seq, chromatin 146. Arabidopsis Genome Initiative. Analysis of the genome sequence of the
immunoprecipitation followed by sequencing (ChIP–seq) and so on. flowering plant Arabidopsis thaliana. Nature 408, 796–815 (2000).
110. Cloonan, N. et al. Stem cell transcriptome profiling via massive-scale mRNA 147. Mouse Genome Sequencing Consortium. Initial sequencing and comparative
sequencing. Nat. Methods 5, 613–619 (2008). analysis of the mouse genome. Nature 420, 520–562 (2002).
111. Lister, R. et al. Highly integrated single-base resolution maps of the epigenome 148. Gibbs, R. A. et al. Genome sequence of the Brown Norway rat yields insights
in Arabidopsis. Cell 133, 523–536 (2008). into mammalian evolution. Nature 428, 493–521 (2004).
112. Mortazavi, A., Williams, B. A., McCue, K., Schaeffer, L. & Wold, B. Mapping and 149. The Chimpanzee Sequencing and Analysis Consortium. Initial sequence of the
quantifying mammalian transcriptomes by RNA-seq. Nat. Methods 5, chimpanzee genome and comparison with the human genome. Nature 437,
621–628 (2008). 69–87 (2005).
113. Nagalakshmi, U. et al. The transcriptional landscape of the yeast genome 150. International Rice Genome Sequencing Project. The map-based sequence of
defined by RNA sequencing. Science 320, 1344–1349 (2008). the rice genome. Nature 436, 793–800 (2005).
114. Wilhelm, B. T. et al. Dynamic repertoire of a eukaryotic transcriptome surveyed 151. Schnable, P. S. et al. The B73 maize genome: complexity, diversity, and
at single-nucleotide resolution. Nature 453, 1239–1243 (2008). dynamics. Science 326, 1112–1115 (2009).
115. Trapnell, C., Pachter, L. & Salzberg, S. L. TopHat: discovering splice junctions 152. Adey, A. et al. The haplotype-resolved genome and epigenome of the
with RNA-seq. Bioinformatics 25, 1105–1111 (2009). aneuploid HeLa cancer cell line. Nature 500, 207–211 (2013).
116. Trapnell, C. et al. Transcript assembly and quantification by RNA-Seq reveals 153. Landry, J. J. et al. The genomic and transcriptomic landscape of a HeLa cell
unannotated transcripts and isoform switching during cell differentiation. line. G3 (Bethesda) 3, 1213–1224 (2013).
Nat. Biotechnol. 28, 511–515 (2010). 154. Howe, K. et al. The zebrafish reference genome sequence and its relationship
117. Johnson, D. S., Mortazavi, A., Myers, R. M. & Wold, B. Genome-wide mapping of to the human genome. Nature 496, 498–503 (2013).
in vivo protein–DNA interactions. Science 316, 1497–1502 (2007). 155. Session, A. M. et al. Genome evolution in the allotetraploid frog Xenopus laevis.
118. Boyle, A. P. et al. High-resolution mapping and characterization of open Nature 538, 336–343 (2016).
chromatin across the genome. Cell 132, 311–322 (2008). 156. Smith, T. F. & Waterman, M. S. Identification of common molecular
119. Ingolia, N. T., Ghaemmaghami, S., Newman, J. R. & Weissman, J. S. Genome- subsequences. J. Mol. Biol. 147, 195–197 (1981).
wide analysis in vivo of translation with nucleotide resolution using ribosome 157. Burge, C. & Karlin, S. Prediction of complete gene structures in human
profiling. Science 324, 218–223 (2009). genomic DNA. J. Mol. Biol. 268, 78–94 (1997).
120. Shendure, J. & Lieberman Aiden, E. The expanding scope of DNA sequencing. 158. Kent, W. J. BLAT—the BLAST-like alignment tool. Genome Res. 12, 656–664
Nat. Biotechnol. 30, 1084–1094 (2012). (2002).
121. Human Microbiome Project Consortium. Structure, function and diversity of 159. Kent, W. J. et al. The human genome browser at UCSC. Genome Res. 12,
the healthy human microbiome. Nature 486, 207–214 (2012). 996–1006 (2002).
122. Tyson, G. W. et al. Community structure and metabolism through 160. Hubbard, T. et al. The Ensembl genome database project. Nucleic Acids Res.
reconstruction of microbial genomes from the environment. Nature 428, 30, 38–41 (2002).
37–43 (2004). 161. Giardine, B. et al. Galaxy: a platform for interactive large-scale genome
123. Venter, J. C. et al. Environmental genome shotgun sequencing of the Sargasso analysis. Genome Res. 15, 1451–1455 (2005).
Sea. Science 304, 66–74 (2004). 162. Butler, J. et al. ALLPATHS: de novo assembly of whole-genome shotgun
124. Blaser, M., Bork, P., Fraser, C., Knight, R. & Wang, J. The microbiome explored: microreads. Genome Res. 18, 810–820 (2008).
recent insights and future challenges. Nat. Rev. Microbiol. 11, 213–217 (2013). 163. Chen, K. et al. BreakDancer: an algorithm for high-resolution mapping of
125. Shokralla, S., Spall, J. L., Gibson, J. F. & Hajibabaei, M. Next-generation genomic structural variation. Nat. Methods 6, 677–681 (2009).
sequencing technologies for environmental DNA research. Mol. Ecol. 21, 164. Ye, K., Schulz, M. H., Long, Q., Apweiler, R. & Ning, Z. Pindel: a pattern growth
1794–1805 (2012). approach to detect break points of large deletions and medium sized
126. Nozaki, H. et al. A 100%-complete sequence reveals unusually simple insertions from paired-end short reads. Bioinformatics 25, 2865–2871 (2009).
genomic features in the hot-spring red alga Cyanidioschyzon merolae. 165. Li, R. et al. De novo assembly of human genomes with massively parallel short
BMC Biol. 5, 28 (2007). read sequencing. Genome Res. 20, 265–272 (2010).
127. Ovchinnikov, S. et al. Protein structure determination using metagenome 166. Robinson, J. T. et al. Integrative genomics viewer. Nat. Biotechnol. 29, 24–26
sequence data. Science 355, 294–298 (2017). (2011).
128. Gymrek, M., McGuire, A. L., Golan, D., Halperin, E. & Erlich, Y. Identifying 167. Chin, C. S. et al. Nonhybrid, finished microbial genome assemblies from
personal genomes by surname inference. Science 339, 321–324 (2013). long-read SMRT sequencing data. Nat. Methods 10, 563–569 (2013).
129. Larsson, C. et al. In situ genotyping individual DNA molecules by target-primed 168. Wang, D. G. et al. Large-scale identification, mapping, and genotyping of
rolling-circle amplification of padlock probes. Nat. Methods 1, 227–232 (2004). single-nucleotide polymorphisms in the human genome. Science 280,
130. Lee, J. H. et al. Highly multiplexed subcellular RNA sequencing in situ. Science 1077–1082 (1998).
343, 1360–1363 (2014). 169. Li, R. et al. The sequence and de novo assembly of the giant panda genome.
131. McKenna, A. et al. Whole-organism lineage tracing by combinatorial and Nature 463, 311–317 (2010).
cumulative genome editing. Science 353, aaf7907 (2016). 170. Kitzman, J. O. et al. Haplotype-resolved genome sequencing of a Gujarati
132. Peikon, I. D. et al. Using high-throughput barcode sequencing to efficiently Indian individual. Nat. Biotechnol. 29, 59–63 (2011).
map connectomes. Nucleic Acids Res. 45, e115 (2017). 171. Fan, H. C., Wang, J., Potanina, A. & Quake, S. R. Whole-genome molecular
133. Shipman, S. L., Nivala, J., Macklis, J. D. & Church, G. M. Molecular recordings haplotyping of single cells. Nat. Biotechnol. 29, 51–57 (2011).
by directed CRISPR spacer acquisition. Science 353, aaf1175 (2016). 172. Seo, J. S. et al. De novo assembly and phasing of a Korean human genome.
134. Zamft, B. M. et al. Measuring cation dependent DNA polymerase fidelity Nature 538, 243–247 (2016).
landscapes by deep sequencing. PLoS ONE 7, e43876 (2012). BLAST and GenBank (GenBank and WGS Statistics; ​https://www.ncbi.nlm.
135. Organick, L. et al. Scaling up DNA data storage and random access retrieval. nih.gov/genbank/statistics/​) were essential tools for sharing and searching
Preprint at https://www.biorxiv.org/content/early/2017/03/07/114553 sequencing data, vastly amplifying the value of each sequence to the field.
(2017).
136. Harrington, L., Alexander, L. T., Knapp, S. & Bayley, H. Pim kinase inhibitors Acknowledgements This is a large topic to cover in a single review. We apologize
evaluated with a single-molecule engineered nanopore sensor. Angew. Chem. to colleagues whose work we were unable to discuss or failed to cite owing to
Int. Edn Engl. 54, 8154–8159 (2015). space constraints. We thank L. Starita, C. Trapnell and A. McKenna for suggestions,
137. Pulcu, G. S., Mikhailova, E., Choi, L. S. & Bayley, H. Continuous observation and T. Tolpa and M. Gillies for extensive assistance with preparing the manuscript.
of the stochastic motion of an individual small-molecule walker.
Nat. Nanotechnol. 10, 76–83 (2015). Author Contributions All authors contributed to the writing of this review.
138. Rodriguez-Larrea, D. & Bayley, H. Protein co-translocational unfolding
depends on the direction of pulling. Nat. Commun. 5, 4841 (2014). Author Information Reprints and permissions information is available at
139. Hyman, E. D. A new method of sequencing DNA. Anal. Biochem. 174, 423–436 www.nature.com/reprints. The authors declare competing financial interests:
(1988). details are available in the online version of the paper. Readers are welcome to
140. Lee, L. G. et al. DNA sequencing with dye-labeled terminators and T7 DNA comment on the online version of the paper. Publisher’s note: Springer Nature
polymerase: effect of dyes and dNTPs on incorporation of dye-terminators remains neutral with regard to jurisdictional claims in published maps and
and probability analysis of termination fragments. Nucleic Acids Res. 20, institutional affiliations. Correspondence and requests for materials should be
2471–2483 (1992). addressed to J.S. ([email protected]).
141. Huang, S. et al. Identifying single bases in a DNA oligomer with electron
tunnelling. Nat. Nanotechnol. 5, 868–873 (2010). Reviewer Information Nature thanks M. Gerstein, S. L. Salzberg and the
142. Rothberg, J. M. et al. An integrated semiconductor device enabling non-optical other anonymous reviewer(s) for their contribution to the peer review of this
genome sequencing. Nature 475, 348–352 (2011). work.

1 9 o c t o b e r 2 0 1 7 | V O L 5 5 0 | N A T U RE | 3 5 3
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
CORRECTIONS & AMENDMENTS
Correction
https://doi.org/10.1038/s41586-019-1120-8

Publisher Correction: DNA

sequencing at 40: past, present and
future
Jay Shendure, Shankar Balasubramanian, George M. Church,
Walter Gilbert, Jane Rogers, Jeffery A. Schloss &
Robert H. Waterston

Correction to: Nature https://doi.org/10.1038/nature24286,

published online 11 October 2017.
In this Review, the year of publication of reference 54 should be 2005,
not 2015. In Box 2, “1982: GenBank (https://www.ncbi.nlm.nih.gov/
genbank/statistics/)” should read “1982: Genbank/ENA/DDBJ” and
“2007: NCBI Short Read Archive” should read “2007: NCBI and ENA
Short Read Archives”; this is because the launches of these American,
European and Japanese databases were coordinated. These errors have
not been corrected.

N A T U R E | www.nature.com/nature