ne w s and v i e w s

Genome sequencing on nanoballs

Gregory J Porreca
Advances in technology deliver cheaper human genome sequencing.

The rapid pace of innovation in the field of performance because the submicron-sized combine to confound a simple bases-per-
genome sequencing continues with a recent DNA features are attached to its surface. So dollar comparison.
publication in Science by Drmanac et al.1. they devised a process to manufacture thin So what can be said about the relative cost
The authors resequenced three full human chambers and perfected a way to flow liquid of this approach? Drmanac et al.1 sequenced
genomes using a next-generation technol- through them, potentially enabling signifi- three genomes at a coverage of 45–87× and at
ogy that combines highly efficient imag- cant cost savings over current systems with an average reagent cost of $4,400 per genome.
ing on ordered arrays with an inexpensive thicker flow-cells. By their estimates, one false-positive sequenc-
ligation-based chemistry. These technologi- Second, throughput is driven both by the ing error occurred every 100,000 bases—an
cal improvements further reduce the cost of speed of the camera and by how many spots accuracy on par with, or better than, that
human genome sequencing. can be packed into a single image. The authors of other popular sequencing platforms.
Next-generation sequencing technologies used the electron-multiplied charge-coupled Complete Genomics, the company associated
© 2010 Nature America, Inc. All rights reserved.

generate up to billions of short reads in a run. device (CCD) present in several other sequenc- with the study by Drmanac et al.1, has posi-
All of these approaches use either polymerase ing systems2 (http://www.polonator.org/), tioned itself as a service provider of full human
or ligase to identify each base with a fluores- which is faster and more sensitive than what genome sequences rather than as a vendor
cent signal that is read by a microscope and a is found in the most popular next-generation of sequencing instruments and reagents. Of
digital camera. Development of these systems platforms on the market. They combined this course, the cost of reagents does not include
has focused on manipulating and arraying camera with one of their key innovations, a equipment, labor and data-handling, and is
DNA such that it can be seen by the camera patterned array of DNA nanoballs. These com- not the same as the price charged to custom-
and sequenced, and on devising a sequenc- pact chains of amplified DNA assemble into ers for a genome sequence. One appropri-
ing chemistry with sufficient accuracy and a densely packed grid of spots on the flow- ate comparison is therefore with Illumina’s
read-length. For effective human genome cell surface, maximizing the yield of useful Personal Genome Sequencing Service, which
sequencing, the individual DNA spots should sequenced bases from camera pixels. Nanoballs delivers a full human genome sequence (on
be small (≤1 µm) and present at high den- offer a higher array packing density than an iMac computer) for $48,000. Complete
sity (approaching 1 million spots per mm2). bridge amplification4, because they physically Genomics currently charges $20,000 for a
Furthermore, the reads must be long enough exclude other DNA molecules from their spot sequence of similar accuracy1,4,7.
(>30 bp) to allow unambiguous alignment to on the grid, and a much easier and cheaper With time, it is certain that prices across
the reference sequence, which is the first step workflow than emulsion PCR5, because they all vendors will fall. Complete Genomics has
in identifying variants. are prepared in a simple reaction that does a target price of $5,000 per genome for ‘bulk’
Drmanac et al.1 have met these goals with not waste most of the amplification reagents orders7, a substantial drop that is certainly
a platform that integrates several technolo- on empty emulsion bubbles. The result is an possible in the near term. Reducing prices
gies: (i) a library-generation protocol that instrument capable of maximal throughput, significantly beyond that will likely require
transforms fragments of genomic DNA into given today’s camera technology, and therefore further innovation. For instance, substantial
highly engineered molecules; (ii) a method minimal capital cost per base pair. increases in instrument throughput could be
for generating spots, called DNA nanoballs, It is difficult to directly compare sequenc- achieved by switching from CCD cameras to
and arraying them in highly dense grids for ing cost between different platforms. This is the much faster and cheaper complementary
efficient imaging; and (iii) a nonprogressive because, for genomic resequencing, cost is metal oxide semiconductor (CMOS) technol-
chemistry (that is, errors do not accumu- driven by the coverage required to achieve ogy. But CMOS is less sensitive, so the DNA
late because each base is read from a fresh the desired accuracy. Different platforms may nanoballs would have to be made brighter,
sequencing primer) that uses ligation with require different levels of coverage to achieve which may require considerable research and
partially degenerate sequencing primers1–3 the same accuracy, so comparisons have to be development.
to yield accurate ~70-bp reads split across made by fixing either coverage, to measure As Complete Genomics makes progress
eight priming sites (Fig. 1). differences in cost and accuracy, or accu- in process automation and robustness, they
What is most intriguing is how the plat- racy, to measure differences in coverage and may be able to address applications beyond
form approaches several technical optima cost6. There are other considerations as well. human genome sequencing, including gene
that in concert drive down cost.  First, the Different mutations (e.g., homozygous ver- expression analysis, chromatin immunopre-
amount of reagent used is dictated by the sus heterozygous substitutions, insertions or cipitation and metagenomics. For these, part
area and height of the instrument’s flow-cell deletions) are generally sequenced with dif- of the difficulty will be in the process scaling
chamber. Drmanac et al.1 recognized that the ferent accuracy. Moreover, the reference stan- and multiplexing required to accommodate
chamber’s height does not affect sequencing dard used to verify mutations must be more the ultra-high throughput of their machines.
accurate than the sequence in question, and In addition, for quantitative applications, it
Gregory J. Porreca is at Good Start Genetics, this is difficult to achieve with the genome- will be important to ensure they can calibrate
Boston, MA. wide single-nucleotide polymorphism chips for biases introduced by the library- and
e-mail: gporreca@gsgenetics.com that are often used. All of these factors nanoball-generation protocols.

nature biotechnology volume 28 number 1 january 2010 43

 ne w s and v i e w s

Drmanac et al.1 Previous approach2 Figure 1 Comparison of the sequencing process

in Drmanac et al.1 (left) and in a previous
sequencing-by-ligation method2 (right).
a Library generation (a) Genomic DNA is converted into library
Library molecules. Each molecule contains four
molecule segments of genomic DNA (i–iv), flanked by
priming sites (shown in purple). Each library
i ii molecule is converted into a linear concatemer
iv iii of itself to become a DNA nanoball. (b) Billions
Rolling circle Emulsion of DNA nanoballs are added to a silicon slide
amplification PCR that contains a grid-like pattern of binding sites,
which causes the nanoballs to self-assemble
DNA nanoballs DNA captured with beads into a dense grid of spots for sequencing,
maximizing the number of useful sequenced
b Arraying bases in each image (see d). (c) Ligation-based
sequencing chemistry is used to interrogate bases
of genomic DNA in the library molecules. Each
cycle of sequencing tags the DNA nanoballs
with a fluorophore whose color identifies the
base (A, C, T or G) present at a specific position.
The chemistry allows 5–10 contiguous bases to
be read from each of the eight priming sites in
© 2010 Nature America, Inc. All rights reserved.

the library molecule. (d) Digital images of the

Patterned array in Random array in patterned arrays are taken after each sequencing
low-volume flowcell larger flow cell reaction. The images are computationally
analyzed to generate billions of raw sequence
c Ligation-based sequencing reads. These reads are then processed with
assembly and analysis software to accurately
Probes Probes identify mutations.
N N N C N N NN N N N N C N N NN N

N N N A N N NN N N N N A N N NN N

N N N T N N NN N N N N T N N NN N
In the span of a few short years, the mature
technology of capillary sequencing has been
N N N GN N NN N N N N GN N NN N
supplanted by new sequencing approaches
that offer tremendous increases in how much
Ligase Ligase we can afford to sequence and how quickly we
Anchor Anchor can do it. As the technology advances, focus
N N NN N N N GN N NN N
G A T C A T T C C G G A A
N N N GN N NN N
G A T C A T T C C G G A A
will shift from the initial feats of sequencing
Genomic DNA Genomic DNA single genomes to the ongoing challenge of
producing lots of sequence accurately and
Degenerate anchors allow Standard anchors allow efficiently. In this endeavour, the platform
62–70 bases per spot 26 bases per spot
of Drmanac et al.1 is sure to remain in the
mix.
d Imaging
COMPETING INTERESTS STATEMENT
The author declares competing financial
interests: details accompany the full-text HTML
version of the paper at http://www.nature.com/
naturebiotechnology/.
Kim Caesar

1. Drmanac, R. et al. Science, published online

doi:10.0026/science.1181498 (5 November 2009).
2009 Nov 5 [Epub ahead of print]
More spots per image Fewer spots per image
2. Shendure, J.A. et al. Science 309, 1728–1732
(2005)
3. Church, G.M. et al. US appl. no. 2007/0207482
As costs continue to drop, the use of and false-negative calls (which generally can- (2007).
sequencing in diagnostics is expected to not be verified) are a source of diagnostic 4. Bentley, D.R. et al. Nature 456, 53–59 (2008).
5. Mckernan, K.J. et al. Genome Res. 19, 1527–1541
increase dramatically. This large market error. Thus, accuracy must be high, quantifi- (2009).
imposes significant requirements on any able and thoroughly measured in advance of 6. Fuller, C.W. et al. Nat. Biotechnol. 27, 1013–1023
(2009).
technology it adopts. Costs must be very low releasing an assay into the clinic. What’s more, 7. Karow, J. Complete genomics details low-cost sequenc-
to displace existing technologies, and accu- continual monitoring of assay performance ing tech in paper; collaborators encouraged by results.
racy must be extremely high. False-positive and compliance with clinical laboratory best- InSequence <http://www.genomeweb.com/sequenc-
ing/complete-genomics-details-low-cost-sequencing-
mutation calls drive up assay cost by requir- practices are imperative if a sequence is to be tech-paper-collaborators-encourage> (10 November
ing expensive and time-intensive verification, considered actionable medical advice. 2009).

44 volume 28 number 1 january 2010 nature biotechnology