Recently I was asked what Roche would purchase when they said publicly that they “would not revisit Illumina, and will pursue smaller takeovers”, and I answered there were a few small development companies out there but even fewer with something ready to sell. (Roche isn’t known for development of NGS as witnessed with their 454 / Curagen acquisition.) And among the firms that was preparing a launch, including GnuBio and Oxford Nanopore, was a small company called Intelligent Bio-Systems.
At ABRF in March (you can view my notes from George Church’s plenary talk here if you haven’t seen it before) Steven Gordon, CEO of Intelligent Bio-Systems, gave a talk during that same session, where he described the technology behind the Mini-20 system they were preparing.
In brief, it has 20 flow-cells that are individually addressable, so it would be 1 flow-cell per individual sample, claim 20M reads per flow-cell (400M reads per fully-loaded run), and an up to 80G sequencing capacity. The run per up-to 4Gb individual flowcell cost is expected to be less than USD $300, and uses a chemistry that is very similar to Illumina’s reversible terminator method. The system was expected to be priced at $120K, competitive to the MiSeq although more expensive than the Ion Torrent PGM. Looking at a simple brochure that AZCO Biotech (a distributor for them) had at ABRF, the readlength is claimed to be 2×55, however the 4Gb per flowcell does not line up with this number. (Calculating 4Gb and 20M reads equates to 200 base-pair readlength.) For now, the ‘up to 4Gb per flowcell’ should be taken lightly, and may be closer to a 2Gb number.
One interesting aspect is that the chemistry depends upon highly diluted terminators – as much as 1:100 labeled to unlabeled nucleotides. Licensed from Jingyue Ju at Columbia University (also a company founder), this is how they are able to lower the cost of the reagents. In terms of template preparation, a polony amplification method is used, and for some applications emulsion PCR is required.
Regarding the cost per gigabase, $300 for up to 4Gb appears to be very attractive. (As of mid-2012, a 1.5Gb run on a MiSeq costs about $1,000, and a 900Mb run on an Ion Torrent PGM is about $750.) However, one of the limitations of the Mini-20 is the turnaround time. Each base addition takes 12 minutes, and thus when you have 200 base-pairs to extend, that is 40 hours. (Although to be fair, that is about the time it would take for the expected 2×250 reads on a MiSeq when the upgrade comes out in a few months.) This holds true for up to 10 samples (since they are individually loaded), however it extends ‘further’ (unspecified) when more than 10 samples are loaded, so there is a penalty to load 20 samples on a single run. This is also independent of the time it takes for template preparation, of which nothing was said regarding how long these processes take.
Since the samples are on a wheel that rotates, one aspect is that a researcher cannot load all 20 samples and walk away; it is a serial process, and loading 20 samples every 15 minutes it doesn’t seem to be that convenient. (That’s five hours of loading samples every quarter hour!) But it does afford the flexibility of loading a sample at any given 15 minute increment.
Looking over my notes and seeing what is on the brochure, there are two key discrepancies, one of which is the cost per flowcell; my notes say $150, the brochure says $300; my notes calculate a 200bp readlength, the brochure says 55bp reads. (I think for this article I will trust the brochure vs. what was verbally said during a presentation.)
IBS was planning early access ‘in Q2′ meaning before the end of June 2012, and commercial launch by the end of 2012. QIAGEN announced this week that they will acquire this company and build a modified instrument based upon this technology and bioinformatics with a partnership with SAP, and launch an instrument in early 2013. (Analysts estimate the acquisition to be on the order of USD $50M.)
When I saw the presentation at ABRF, and looking at what the Mini-20 offered (IBS also developed a larger system called the ‘Max-Seq’, but I am not aware of how many they sold of those through AZCO Biotech), there was a realization that this company had a long road ahead to build out the commercial infrastructure to sell and support this instrumentation. With QIAGEN, those pieces are already in place, and QIAGEN plans to include sample-preparation automation (purification up-front of the sequencing of DNA and RNA) in addition to automation of the library sample preparation (which it has not done before but several liquid handling companies, notably Beckman, Caliper and Tecan, have already done).
There are several unknowns that QIAGEN will have to solve for in bringing the Mini-20 to market. (I am referring to the research market here, although QIAGEN will likely straddle the translational / clinical market as well.) First is the data quality of the chemistry itself, as the market while familiar with the approach has been conditioned by vendors to take all claims with a skeptical eye. (The AZCO brochure claims ’82% of reads up to 55bp are >Q30′.) Second would be to automate the template preparation (polony formation) as customers won’t accept much manual processing to generate sequence. Third would be to have an open informatics pipeline, something that QIAGEN may be reluctant to do (speaking from personal experience of the company’s conservative culture). Without an open pipeline, the ‘black box’ approach may work in a diagnostic or clinical setting, but is not acceptable for the research customer, as parameters and tools are almost infinitely variable.
QIAGEN has a long road ahead of itself to launch this new platform, and will face stiff headwinds in 2013 should Oxford Nanopore or Genia start to deliver upon the promise of nanopore sequencing. Of course, simultaneously Illumina and Ion Torrent won’t stay still in delivering better and faster and cheaper, so the competition will only intensify.