Illumina's Blizzard

As the clear leader in the genome sequencing technology market, Illumina is often in the news.  As befitting the winter season, there's just been a blizzard of Illumina news, mostly announcements from within but also two important external ones (an excellent summary of these can be found at NextGenSeek).

Two stories that often dominated last year's genomics business news was the attempt by Roche to buy out Illumina and Illumina's attempt to woo Complete Genomics away from BGI.  Each dragged out for an extended period, with the Complete Genomics saga going through a particularly convoluted path, capped by the nearly absurd claim that BGI acquiring Complete would have serious U.S. national security implications (yet somehow Illumina continuing to sell BGI an other Chinese firms their instruments did not).  A third suitor also appeared, initially identified only as 'J' but later curiously revealed to be an investment firm.  Curious in the sense that it was not Life Tech (which is busy with Ion and letting SOLiD languish -- and never gave StarLight a serious outing) nor one of the other big genomics players (such as Agilent, Roche, Perkin Elmer or even Affymetrix) nor a possible medical powerhouse wanting to break into genomics in a big way  (such as GE or Siemens).  All that appears to be at an end; BGI has received the necessary clearances for foreign acquisition of Complete and Illumina has withdrawn their bid.  

On the Roche front, the head of Roche has emphatically denied continuing interest in Illumina, scotching the rumors of a new bid that surfaced at the end of the year.  I have yet to encounter someone in sequencing who viewed a Roche acquisition as a good thing; the consensus is that Roche let 454 wither on the vine through a lack of investment.  Indeed, as suggested by Matthew Herper, if Illumina had been acquired by Roche the only good outcome from the standpoint of the sequencing community would have been if Illumina's management effectively took over Roche.

Illumina's business development office has stayed busy on other fronts.  The week started with the acquisition of Verinata, a company developing prenatal diagnostics by looking at fetal DNA circulating in the maternal bloodstream.  This is potentially a huge market, though with several other players and a complex IP landscape (lawsuits already flying).  A second acquisition was announced later in the week, of sample preparation company Moleculo.  Moleculo has developed a process for sequencing size-selected DNA with short reads in a manner which enables reconstructing the individual fragments from the input material, generating very high confidence synthetic sequences.  The big potential market for this is phasing SNPs, but it is likely to also make a difference in de novo sequence assembly (I've tried it on hard samples; it works, though it's not the magic bullet I need).  

Illumina also launched new Nextera-based products, including a Nextera targeted capture solution and a really cool mate pair kit, proving that their Epicentre purchase is a gift that keeps on giving.  Badly overdue (it had been once slated for summer of last year; see poster on page 20), it has a number of several nifty properties.  Input DNA can be as little as a microgram, in contrast to the tens of micrograms demanded by many existing protocols.  Two different workflows are available.  The gel-free method gives very high complexity libraries from smaller sample amounts and a simpler workflow, albeit with a wide distribution of jump sizes.  This is claimed to enable high quality microbial assemblies from a single library; no need to run a separate short insert paired end library.   Alternatively, a slightly more involved gel-purification workflow can be used to generate less complex libraries with a tight size distribution.   The process (detailed on page 20 of this slide deck) uses transposomes to both fragment the DNA and biotin tag it.  Circularization is by ligation; some recent mate pair schemes have used recombinases to circularize the DNA.  Conventional mechanical fragmentation then breaks the circles, the optional size-selection occurs and the biotinylated pieces are captured and a standard TruSeq library preparation follows.  Would have been cooler if these steps were also Nextera-based, but cool isn't always the way to go.

Illumina also repeated their promises that this summer will see 2x300 runs on MiSeq yielding 15Gb.  HiSeq 2500 rapid runs will increase to 2x250 bp with 300Gb of data.  Also exciting is the announcement of a new flowcell scheme which apparently generates ordered arrays of clusters (as opposed to the random arrangement found now), enabling a doubling of the number of useful clusters.  All of these should enable Illumina to maintain its dominance of the sequencing research market, while also helping set up the bigger payoff clinical applications such as those developed by Verinata.  The speed of HiSeq rapid mode should largely hold off Ion Proton (when it finally reaches a human genome per run), and may well make Complete Genomics technology nearly obsolete in the clinical space, given that "actionable medical result" and "needed stat" routinely show up in the same sentence.

Illumina's acquisition of Moleculo also mirrors the Epicentre acquisition by boxing out the competition; neither of these technologies is really Illumina-specific, but now Illumina has locked them up.  For the pretenders to the short read sequencing throne (Life Tech, Qiagen), novel sample preparation solutions might be an path to the palace. If I were at either, I'd be beating hard on BGI/Complete's door to get access to the Long Fragment Read technology, as this is potentially a winner.  Similarly, Life would be wise to see if their Wildfire/Avalanche technologies can accommodate the recently published scheme for on-flowcell library preparation to generate long-range sequence information.  Interest in generating long-range information (haplotypes) is heating up; it would be unfortunate for the other companies (and probably for the sequencing community) if Illumina monopolizes this space.  Not every novel sample preparation method has legs (anybody Long March-ed a sample recently?), but the real gems can dominate.  In particular, Nextera has proven to be a rich vein, and I suspect (given my own flights of fancy scheming) there will be many more nuggets to be found.

Long term, the real threat to Illumina is not another short read technology, but super long read methods.  However, there is little reason to think that Illumina will have serious competition here.  PacBio presented their 2013 roadmap this week, which promises interesting improvements in yield and read length: 1Gb per SMRTcell and 9Kb+ reads.  Still, I expect PacBio to remain a niche instrument, valuable for microbial genome assembly and methylation studies, but still not cost-efficient for widespread use for mammalian epigenetics.  Genia has pushed-off commercialization until 2013, using a new chemistry.  Oxford Nanopore

remains unencumbered by public datasets; their sole announcement in the new year is around nailing down more IP.  Perhaps AGBT will provide some fireworks (and for once, I'll be on site for the sparklers or squibs), but for the moment it looks like Illumina will retain its crown.