Monday, 10 November 2014

HiSeq X Ten: when might they be available one at a time?

In the Summer I posted a Christmas letter to Santa. He's already delivered number 2 (cheaper RNA-seq) and 3 (longer RNA-seq reads via PE250), and number 4 might be coming soon (exomes at PE125). I'd also asked that he not deliver HiSeq X Ten as an individual instrument just yet, but as there are just 44 days left till Christmas I thought I'd look head and see what might be the reasons for, or not for gift wrapping a single X Ten this Christmas.

The HiSeq X Ten is an awesome system capable of sequencing around 18,000 genomes per year; you'd need thousands of circa 2007 GAI's to do the same job, and 30 HiSeq 2500's. With a ratio of three to one many labs running HiSeq 2500 would love to switch over to X Ten, they'd get faster and cheaper genomes, more lab space and fewer runs to perform for the same output, but Illumina do not sell X Ten one at a time - yet.

What's stopping Illumina: The sales of X Ten systems (10 X Ten's) seemed to blow away Illumina's expectations so they probably have enough demand to satisfy themselves for now. They would need to confirm X Ten could be used for more than Human genomes, as an X Ten would not be much use to a lab like mine as most of our work is RNA-seq, exomes, amplicomes and ChIP-seq. Also Illumina would need to be able to provide a reasonable upgrade path to HiSeq 2500 owners, many of whom have just forked out for V4 upgrades or new 2500s. 

What about HiSeq 2500: It's pretty obvious to users that the V4 2500's are fantastic machines capable of delivering lots high quality of data in a very reasonable time frame and at somewhat reasonable cost. The X Ten must have been developed on HiSeq 2500 instruments that were hacked by the R&D team. This could mean that an upgrade might be a possibility.

Illumina now have three quite different clustering chemistries (X Ten, 2500/MiSeq and NextSeq) and two SBS kits to maintain (X Ten/2500/MiSeq and NextSeq). I am not clear as to how different some of these are to each other, but too much difference e.g. HiSeq X Ten vs 2500 clustering, or HiSeq vs NextSeq SBS must be a headache to manage. Perhaps we'll continue to see Formula 1 style trickle down of improvements from one system to another as we did with the most recent 2500 RTA which appeared to be ported over from the MiSeq and now allows much lower complexity libraries to be sequenced.

In fact if Illumina could give users ordered flowcells on HiSeq 2500, and that can be used for any library prep type then they'd probably satisfy most of us!

So will we be able to buy an X Ten after the 33rd J.P. Morgan healthcare conference in Jan 2015? I'm not sure. One thing can be said, that Illumina cannot be accused of resting on their laurels, so anything is possible. X Ten is about to get PCR-free genomes so RNA-seq or exomes could be on the agenda next year, and many X Tens are running at well under the 18,000 genomes per year capacity. To recoup the investment into X Ten development Illumina need to ship sequencing kits, and lots of them. Perhaps the thing that might bring X Ten to the masses is the inability of large centres to fill their machines?



    Take a look at the part on the NextSeq v2 chemistry. I think it is possible that Illumina would like to transition over to emphasizing the NextSeq (and successors) as an instrument of choice given all the maintenance issues on the HiSeq 2500s.

    Already the high output NextSeq mode can give substantially more data than a HiSeq RR -- the question is how far they can scale the technology so that it can be on par with a HiSeq TSv3 or HSv4 run. I think that adapting ordered flowcells into the NextSeq is in the cards as it should be independent of the specific SBS chemistry.

    The issue is also at what rate do they want to start migrating technology platforms. The HiSeq X is just a year old, and academic institutions are generally constrained on capital expenditures... although I'm sure that there's a lot of pent up demand to start replacing the older HiSeq 2500s that are getting long in the tooth....

    1. Interesting ideas. However, I don't think the addition of patterned flow cells to the NextSeq is likely to happen soon. The NextSeq flow cell is HUGE, primarily to make the optics easier - the individual clusters are large and far apart, making them easy to distinguish. Patterned flow cells would necessitate much better (and more expensive) optics. Not that the NextSeq won't migrate to this, but it may be a while.

    2. I wouldn't say that the clusters are necessarily farther apart on the NextSeq (judging by the thumbnails), but they are much larger. I think patterned flowcells are such a natural followup because they get rid of a lot of computational complexity and the cluster density can be directly manipulated through the pattern density.

      I have almost no doubt that the NextSeq HO mode can hit TruSeq v3 densities with or without the patterned flowcell. I don't think they would have released the platform if they couldn't scale it down the road. I think most people would be quite happy if the NextSeq could hit TSv3 outputs.

  2. " to start replacing the older HiSeq 2500s that are getting long in the tooth...." ... and are starting to suffer from mysterious old age problems