Monday, February 18, 2013

AGBT Preview: Nabsys

A complaint which seems to be circulating on Twitter and elsewhere is that this year’s AGBT conference on Marco Island next week doesn’t look like it will have any excitement around new platforms.  AGBT has been a traditional coming out party for platforms.  Last year it was Oxford Nanopore which created a huge buzz, and in previous years that crown has been held by Ion Torrent, Pacific Biosciences, Complete Genomics and others (including a few which seem to have gone kaput).

It is hard to argue that this year’s program is much more heavily tilted towards applications of genomics than novel genomic technologies.  Many of the genomic technology talks are updates on the evolution of existing platforms such as PacBio and Illumina (especially the Moleculo technology).  But, there will also be novel technologies.  

This past fall I had the pleasure of spending half a day with the folks at Nabsys, located a short distance from Boston in Providence RI.  Nabsys’s buildings are in a sea of drab parking lots (not to mention sheriff’s cars; a courthouse is nearby), but on the interior were quite pleasant.  I was largely wearing my “day job” hat that day, but Nabsys will be unveiling their instrument at AGBT and has given me permission to talk about what I saw (and reviewed what I've written to make sure I didn't make some dumb errors, though the content is all mine).  Of course, I will update this after I look some more at AGBT.


It is worth noting that if Nabsys can remain on track for commercialization, with a beta program imminent and commercial launch late in the year, then they are likely to deliver the first molecular biological analytic device relying on an electronic, single-molecule detector. While a lot of excitement has been focused on nanopore companies such as Oxford Nanopore and Genia (and others), perhaps it is the difference in goals which has enabled Nabsys to approach the finish line sooner.  Rather than attempting to actually sequence DNA, Nabsys is promising “Positional Sequencing”, which really means precisely measuring the distance between predictable landmarks on DNA.  

Positional sequencing is in some ways provides similar information to the optical mapping systems of OpGen and BioNano Genomics, but with two important twists.  First, by employing electronic rather than optical detection, Nabsys can obtain much more accurate measurements of position. Second, while the optical mapping companies need nucleases to define their landmarks, Nabsys’ are defined by oligonucleotide probes, and hence the resolution of a map can be tuned by tuning the complexity of that probe pool.  Ultimately, the goals is to zoom all the way down to individual nucleotides (effectively), but this ability to dial the resolution means that Nabsys can follow a classic “disruptive innovation” trajectory: deliver a tool that is initially useful, inexpensive but perhaps a little crude.  But, a clear path to performance improvement exists.   A challenge, of course, is whether Nabsys can carve out a foothold, and drive to that final resolution, with the spectre of one of the straight nanopore sequencing companies bringing their product to market.

The initial focus application for such a device will be small genome scaffolding (a topic VERY dear to my heart).  This has been a key market for OpGen and likely one for BioNano.  A key difference is that these systems cost many fold more than the approximately $50K proposed for the Nabsys box.  An obvious future market is structural variation mapping, though one challenge is that some of the most exciting places to apply such analysis are cancer samples, which too often have been through the harsh formalin fixation process (and embedding in paraffin wax) which tends to shred DNA into tiny pieces.

During my visit, I saw alpha instruments (though alas, did not see them run), which was an ovoid structure that my hand could have just fit over.  They were mounted on a Hamilton robot, though that was largely for convenience and they are actually designed to be manual operated as well.  A modest number of reagent additions are necessary during a run.  Sample preparation is relatively simple: sequence-specific sites are labeled with probes and a proprietary reagent mix gets rid of the biophysical problems with long DNA molecules (the analogy Nabsys used is taking cooked spaghetti and turning it into uncooked spaghetti).  Probes are labeled in a proprietary manner so that they can be distinguished.  An instrument run involves driving the DNA molecules one at a time (but at high rates) through solid-state detectors, which enables their labels to be read. Molecules move through each detector at a speed of 1 Mb per second.  As with other purely solid state semiconductor technologies, a clear path exists to scale the number of detectors per chip, promising regular performance improvements.  

As with any new technology, a key question will be “when doesn’t it work?”.  Large single-stranded DNAs are something to make many people nervous, as complex secondary structures could form.  If these interfere with probe binding or passage of the DNA through the detector, then that could be a serious problem.  The sensitivity of the system to DNA purity will be another important parameter; recent in house work has illustrated the challenge of routinely generating very high molecular weight DNA (think 100s of kilobases or more) free of cellular debris, particularly if the menagerie you are trying to characterize are phylogenetically diverse.  Human samples can similarly be challenging.

Nabsys will be giving more details at AGBT, and I’m looking forward to another in depth technical discussion with them (though, regrettably, the lovable Cavalier King Charles Spaniel that patrols their office-only building is unlikely to be in attendance).  I’m hoping I can get some photos to post; I was so excited to actually hold a solid state DNA detector in my hand last time I forgot to ask if I could shoot some!    Still, as someone who is unsatisfied with the solutions I’ve tried so far for genome scaffolding, I’m itching to try it out.  They'll be running real samples in their AGBT suite, though (alas!) it's not BYOD (Nabsys is bringing the material), which lends credibility to their plans.

6 comments:

Anonymous said...

Sorry but I can't see how this can compete with single molecule nanopore sequencing at any time

Paul T Morrison said...

Thanks Keith, illuminating as always. "When doesn't is work" is always a useful question. My question would be what projects would this first iteration be useful for? My guess would be de novo assembly of our favorite animals Streptomyces and e. coli. I still marvel at how all current technologies have such an inability to spit out one contig with the depth and length of fragments available. This little box could help scaffold them.

Heres another one. Say you know there is a certain genomic arrangement that brings death and destruction. A bar code reading the hardened spaghetti of this awful bacteria or virus could be read by Nabsys.

I guess I can wait a couple of days before I seek answers to a bunch more dumb questions that are coming to mind.

As for Anonymous' question, my response would be, "have you seen one nucleotide from single molecule nanopore?" I hope we see some soon and they are not all CACA.

Dan Koboldt said...

Keith, great post! You have a talent for getting the exclusive look at things. I look forward to hearing what Nabsys shows off at AGBT.

Anonymous said...

Hi Paul,

Yes, and on this blog site

http://omicsomics.blogspot.com.au/2012/02/oxford-nanopore-doesnt-disappoint.html

Paul Morrison said...

No Anon, I read that post a year ago. Great post. But I am still waiting for a few nucleotides of data.

Keith Robison said...

Anon: Yes, I've acknowledged being excessively exuberant about Oxford Nanopore last year. Given their pedigree, they certainly can't be reduced only to the butt of silly jokes about their one product's name (not that I don't frequently do so on Twitter).

Still, as Paul points out, until nanopores start generating data & we can really pit systems head-to-head with all the various scenarios, there aren't clear winners and especially not clear take-outs.