James Clarke, Hai-Chen Wu, Lakmal Jayasinghe, Alpesh Patel, Stuart Reid, Hagan Bayley (2009). Continuous base identification for single-molecule nanopore DNA sequencing Nature Nanotechnology DOI: 10.1038/nnano.2009.12
The clever boys and girls at Oxford Nanopore Technologies - one of the most quietly impressive contenders in the hotly-contested next-generation DNA sequencing race - have a new paper out in Nature Nanotechnology today. The paper demonstrates proof of principle for a crucial step in their approach to DNA sequencing, the accurate recognition of DNA bases as they pass through a tiny protein nanopore.
Oxford's approach is outlined in cartoon format in this video:
(See the Oxford Nanopore website for a better-quality Flash version of the video.)
Put simply, the system works by sequentially chewing DNA bases off the end of a long strand and then detecting each cleaved base as it falls through a protein nanopore. In today's paper the company demonstrates the use of engineered nanopores to achieve accurate recognition of five different DNA bases (the standard A, C, G and T, as well as methylated C).
Here's the crucial figure:
The picture above shows the traces left by the four different DNA bases due to the different effects these molecules have on electrical current flowing across the nanopore. Although there's a fair bit of noise at the molecular level, these signals provide surprisingly accurate base detection: the histogram on the left shows the fraction of bases correctly assigned (base-calling errors fall in the troughs between the four peaks). The average accuracy is 99.8%, and importantly the errors aren't random: where there is ambiguity, there are only two possible states (rather than four). That type of error constraint will make it much easier for downstream analyses to minimise the impact of base-calling errors.
There's still plenty of work to be done before this technology can be converted into a commercial platform, but I understand that the company has made substantial progress since the work described in the paper was completed over six months ago. Clearly Oxford is doing something right - the recent $18 million cash injection from sequencing giant Illumina, with a promise of more funding to come if Oxford met specific (unnamed) technical benchmarks, was an impressive vote of confidence.
Oxford's technology strikes me as the most exciting in the emerging third-gen arena: the nanopore approach offers some hefty potential advantages over competing technologies.
Firstly, the platform detects DNA directly rather than relying on
indirect labelling steps, which is important for several reasons: it
means that sample preparation should be trivial (essentially just DNA fragmentation), reagents will be much cheaper, and there will be no need for amplification of the input DNA - eliminating a major potential source of bias.
Secondly, this paper demonstrates that the system can detect the modified DNA base methyl-cytosine directly,
which will be extremely useful for epigenetics (the analysis of
physical modifications of DNA, which can have effects on the levels of
gene expression). Current systems for detecting methyl-cytosine require
a fiddly bisulphite substitution step; direct reading of modified bases
would be a huge advantage.
Thirdly, if the company can combine their nanopore reader with a DNA-cleaving enzyme with sufficiently high processivity, the platform would potentially offer extremely long read lengths.
The ability to sequence single molecules of DNA with high accuracy over
thousands of base pairs would make assembling whole genomes much easier
and far more accurate than is the case for current platforms (which
have read lengths ranging from around 35 to 500 bases).
Finally, the simplicity of the system's readout as electrical signal means that downstream data processing and storage are easier, and that the system is readily scaleable - manufacturing techniques for handling high-density electrical sensors are already very well-established. That means creating a massively parallel array of nanopores to sequence huge numbers of DNA molecules simultaneously should be relatively straightforward.
Oxford will be up against a strong rival in Pacific BioSciences, who look set to launch their own commercial long-read single-molecule sequencing platform in 2010, but the advantages of nanopore technology and their backing from Illumina make the UK-based company a serious contender. I'm looking forward to seeing how this field evolves in 2009.
Media coverage: So far, a nice piece by Kevin Davies in Bio-IT World and a story in the Times focusing on the broader implications of cheap sequencing technology. A new blog, Biotechnically Speaking, focuses on the implications of this technology for epigenetic research.
While I think that Oxford Nanopore are on the way towards a great sequencing product, I still prefer the Pacific BioSciences approach from a biologists point of view. Somehow, using a polymerase for sequencing has always seemed like the more "natural" way to go. And since Pacific's method doesn't mess with the DNA, you would be able to work with that DNA after the sequencing run, something that is not possible after Oxford's exonuclease has chewed away the nucleotides. On the other hand, being able to get methyl-cytosine directly in the sequence rocks!
I don't agree with Argent23. Yes, it would be good to have a '3rd gen 'method that allowed reuse of the DNA but I am not sure that PacBio is that method. According to their Science paper they must be accumulating DNA photodamage, especially to their G bases. This is why the system ultimately 'grinds to a halt' after a certain amount of exposure of the sample to fluorescence. Unfortunately, we have seen the same photodamage in all optics-based sequencing methods.
I agree with two previous comments that both Oxford Nanopore and PacBio have demonstrated proof of principle for single-molecule sequencing. However it is my opinion that useful technology will be based on physical principles of reading DNA bases in nanopore. Indeed use of chemical and enzymatic reactions results in intrinsic inevitable high rate of errors of sequence data, that was proven by both publications of Oxford Nanopore and PacBio in Nature and Science. In addition use of fluorescence has its own fundamental rate of errors due to low photon yield from single molecule. Therefore I believe that only Single Molecule Raman which currently exceeds by brightness fluorescence 10000x times will be way to read DNA in future Nanopore based sequencer with speed up to 1 Mln base per second
Interesting, but fundamentally flawed.
Care to elaborate?
I sorry I did not realized that comments of "2b1" could relate to my posting. Just in case if "2b1" comments relates to my posting and not to Oxford Nanopore I am responding to 2b1 concerns. At fundamental level Raman read approach is OK . It is well known now for 10+ years that at single molecule level plasmonic enhanced Raman beats substantially fluorescence by brightness of signal even in not optimized case.
I would advice to "2b1" to read papers
6.S. Emory, S. Nie, âProbing single molecules and single nanoparticles by surface enhanced Raman scatteringâ, Science, 275, 1102 -1106 (1997)
7.Kneipp, K., Y. Wang, et al. (1997). "Single Molecule Detection Using Surface-Enhanced Raman Scattering (SERS)â, Physical Review Letters 78(9): 1667
8.A. Michaels, L. Brus et al., âSurface Enhanced Raman Spectroscopy of Individual Rhodamine 6 G Molecule on Large Ag Nanocrystalsâ, J. Am. Chem. Soc., 121, 9932 (1999)
Just wanted to bring a dose of reality here: they have not, and no one ever has, sequenced DNA of any length by this method. What the "crucial" figure "a" shows is a bunch of random nucleotide monophosphates being detected as they come through the pore- that's all, no sequence data. They're actually a long way off, in fact... read the paper.