Name: Stuart Lindsay
Age: 57
Title: Professor of physics and chemistry (since 2003) and director, Center for Single Molecule Biophysics (since 2002), Biodesign Institute, Arizona State University, Tempe
Experience and Education:
Assistant/associate/full professor, Arizona State University, since 1979
Co-founder, Molecular Imaging (now part of Agilent Technologies), 1994
Consultant, Philips Industries, London, 1977-79
Research fellow, University of Manchester, UK, 1975-77
PhD in physics, University of Manchester, 1976
Diploma in advanced studies, University of Manchester, 1973
BSc in physics, University of Manchester, 1972
Stuart Lindsay, director of the Center for Single Molecule Biophysics at the Biodesign Institute of Arizona State University, has been working on a nanopore-based approach to DNA sequencing, called "sequencing by recognition," that uses hydrogen-bond-mediated electron tunneling to recognize the bases.
Earlier this year, his lab published two papers — one in Nanotechnology, another in Nature Nanotechnology— showing that they can read the base composition of DNA by electron tunneling. His group has another paper in press that demonstrates single-base resolution.
In Sequence spoke with Lindsay last week about his research, and what kind of DNA sequencer might eventually come from his lab's work.
Could you explain how 'sequencing by recognition' works?
The basic technology here is nanopore sequencing. One of the key missing ingredients in nanopore sequencing is a chemically selective readout. The problem is that you need the readout to be localized to a single base.
One of the proposals has been electron tunneling, which is a very localized effect. In essence, it's the process whereby atoms share electrons when they form a chemical bond, so it's restricted to distances of angstroms. Therefore, perhaps you could tell which base is passing through a nanopore by recording the electron tunneling current through the base. That proposal has all kinds of problems with it by itself. One of the things that my lab has worked on for many years has been tunneling through single molecules, and making reliable contacts with the molecules is 99 percent of the problem.
So what we set out to do is combine two elements: one is electron tunneling to give you an electric signal of a base, and the second is chemical recognition to specifically recognize the base. You are probably aware of the beautiful work that Oxford Nanopore [Technologies] and Hagan Bayley [at the University of Oxford] have done on chemical recognition for detecting nucleotides. As a reformed physicist, I love chemical recognition, and I think that Hagan's approach is great. What we would like to do is combine chemical recognition with electron tunneling.
In our Nanotechnology paper, there is a beautiful picture of a probe with a base attached to it recognizing a target base on a single-stranded DNA that is connected, to make an electrical circuit, at the backend by guanidinium ions hydrogen bonding the phosphate. Thus, you complete an electron tunneling path through the circuit, and the circuit goes, 'Bingo, I recognize my complementary base.' That paper actually showed that 'yes, this works,' in the sense that it did detect bases with unmodified, unlabeled DNA, and did so with quantitative accuracy. However, it did not resolve single bases.
[ pagebreak ]
The Nature Nanotechnology paper is actually the physics behind what was going on in that measurement, and explains why we did not resolve single bases. It describes the physics of the tunneling junction, the fact that in that geometry, the path length for electrons to tunnel through is just so large. The short-term bad news from that is that our original proposal is not going to work. The good news is that if you understand that the problem is just one of limited signal going across the whole DNA, there are plenty of other tunneling paths that one can use that are shorter, so that one can then generate a chemical recognition signal.
In fact, following on from the Nature Nanotechnology paper is a paper in press in Nanotechnology, that gives the quantitative data for tunneling through a single nucleoside pair. That paper is just about to appear, and it supports the calculations that we are then using as the basis to close the loop and design what should be a reliable tunneling readout system.
What do you still need to do to make this useful for DNA sequencing?
There are a couple of steps. Having identified the key problem, we have designed and made reagents that will make a shorter tunneling path through a DNA base. And then, the absolute key step, of course, is to build this thing. And you need to appreciate that to actually make something like that is non-trivial. But we do have some very new technology that is extremely promising for making tunnel gaps self-align with nanopores, and that is what we are working on right now, so watch this space.
What are the potential capabilities of a DNA sequencer that is based on this technology, if it is going to work, compared to currently available DNA sequencers?
The first two important implications are common to all nanopore sequencing technologies, which are, first of all, the possibility of [read] lengths that begin to approach genomic lengths. It takes very complex measurements of tiling arrays right now to uncode long-range genomic complexity [like] inversions, delections, or repeated genes. Would it not be fabulous if you could just [read] the genome and spit out the true sequence with minimal assembly?
The bad thing about our proposal, because we use specific chemical recognition, is that unless we are lucky enough to have the circumstance that Hagan Bayley has, which is that one chemical recognition element will do four bases, we have to rely on four different DNA readers, [so] we are still going to be doing some sequence assembly. But sequence assembly based, presumably, on hundreds or thousands of kilobases of DNA.
The second advantage is that it's label-free and reagent-free, other than the reagents that are used to make the device itself. Dan Branton, in a review in Nature Biotechnology, came up with a figure of $40 in preparation cost per genome. The point is that the chemical preparation cost could be very low.
The sequencing speed is quite easy to predict, because for all sorts of technical reasons, it's hard to go much faster than about a base per millisecond. One could maybe to go 10 times faster than that. That means sort of a genome in a month with one reader, so highly parallel technology would be required. It's probably buildable to bring it down to very short sequencing times.
The other advantage is that all of the single-molecule sequencing techniques are just that, so that it might become possible, eventually, to make sequencing so cheap and so undemanding of materials that one could look at genomes from cells one has selected, for example neoplastic cells. It might be conceivable that a cell-by-cell genomic analysis could eventually be enabled.
Then there is a final thing, and Hagan Bayley has already demonstrated this. If you are doing artificial chemical recognition — in other words, if you are not doing a polymerase-based sequencing technique — you don't have to confine yourself to looking for natural bases. For example, if methylation is something you would like to read out, that's something you can design into your chemical recognition. It's obviously a way off, but there isn't a physics or chemistry reason, in principle, why one could not do that.
What about the raw accuracy of such a device?
That's an interesting question, and I don't know how well one can answer that. If you look at Hagan Bayley's discrimination, based on trapping times in cyclodextrin, the accuracy is actually rather impressive. I think that depends a lot on how the recognition reagents are designed. In some of the preliminary work that you see in our papers, the recognition efficiency is something like about 80 percent, without tuning the system. Of course, to get to 99.999 percent [accuracy], one is going to want multiple reads, but on the other hand, I think there is a lot of scope, once one understands the basic chemistry and physics, for improving the efficiency of recognition.
Are you already thinking about how this might be commercialized in the future, and do you have any partnerships for that?
I wish. This is the world's worst market for doing anything that's way out there. We do have intellectual property, and we have talked to some people, but … that's about all I can say.