Saturday, February 28, 2015

What's Been Cooking for Ion At AGBT15

Rounding out my remote coverage of platform news from AGBT, the Ion Torrent team also lent me some of their time (and at risk of sounding obsequious, I do greatly appreciate this -- vendors have almost no down time at these events) to touch on some of the topics I I wrote about in my Ion history and speculation piece.

We set the table by discussing where Ion sees its core markets; the team is realistic that with the current hardware lineups they won't be competing with Illumina for whole genome sequencing of large genomes.  But Ion believes that for targeted resequencing, particularly in the oncology space, that they have cooked up solutions that are competitive with those from the San Diego titan. I had remarked in my piece that Illumina had gobbled strategically to bolster their platform, and the Ion team reminded me that Life Technologies had acquired Compendia Biosciences, which had developed a large knowledgebase (OncoMine) in the oncogenomics space.  The Ion team has used this resource both to develop panels focused on high value (frequent, actionable) mutations and fusions, but also to assist users with interpreting results, particularly within the context of clinical trials. The team assured me also that the level of investment in the platform now under Thermo Fisher as they had under Life Technologies; Ion remains a very important piece of that complex pudding.


Ion is very confident that their targeted AmpliSeq is cost-competitive with microarrays for expression or marker profiling large numbers of samples, and that with the release of the PII chip later this year may even offer a significant cost advantage (more on that chip below).  Off-the-shelf panels are available only for human, but with AmpliSeq Designer a researcher can create their own panels. 16S profiling of microbiomes and exome sequencing were mentioned as markets with growth potential for the AmpliSeq line. For expression profiling, targeted AmpliSeq amplicons are designed to a single common exon-exon junction per gene, with 21K amplicons per reaction.

Ion is releasing at AGBT data showing AmpliSeq with only 100 picogram input which shows similar performance statistics (such as coverage) with 10 nanogram input.  Since 100pg is fewer than 30 haploid copies of a human genome, that's pretty amazing and probably hard to beat without some sort of WGA or other pre-amplification. In the clinical space, particularly oncology, requiring such tiny nibbles of tissue can be highly enabling. Even more exciting is that Ion is developing a next generation of AmpliSeq products will be able to work directly on select biological inputs such as buccal cells or blood, banishing a separate DNA preparation step in these cases.

Isothermal Template Preparation

I've been intrigued by the isothermal template preparation chemistry for some time, so we also discussed its integration into Ion's mise en place. Ion has three main commercialization efforts for this technology running in parallel.  For HLA typing, a single tube format of the template prep is being used; this is available as early access. A second is very rapid pre-implantation screening of copy number variants for the in vitro fertilization market, also in single tube format.  The third is to run the chemistry on the Ion Chef platform.

I got a good lesson in the advantages and trade-offs of the isothermal chemistry.  Running it in a single tube doesn't require investing in a Chef ($55K) and is significantly faster, but has lower sequencing yields because the partitioning into amplification compartments is not as stringent (hence more compartments receive multiple templates and yield a useless bead).  Isothermal amplification can more efficiently amplify long templates, opening the possibility of 600 basepair reads on the PGM chips and 400 basepairs on the Proton PI. Tech access to these meso reads (if your reads are many kilobases long, I can't call them long reads!) on the Proton in the second half of the year.

I asked about read length constraints: it's both at the level of chemistry and phasing algorithms.  Signal-to-noise is (not surprisingly) the key attribute, which explains the superior read lengths on the larger bead PGM chips -- these have more signal since they have greater surface area.  Ion believes they have solved well-to-well cross-talk; their focus has been improving chip coatings and bead surface chemistry to enhance the amount of signal which can be detected in each well.


As mentioned in my prior piece, the Proton PII chip has had a difficult launch, with previous versions sent back to the kitchen for a restart.  At last year's AGBT, Ion discussed significant improvements in the chemistry, but difficulty extracting data off the chip with sufficient rapidity (5Gbits/s sounds fast, but wasn't fast enough): only quadrants of that design could be read out.  After a complete redesign of the chip, they are now past that problem.  These new PII designs were received in mid-January and have demonstrated full chip access, with 200-300M reads of 100 basepairs.  Ion sees this performance as competitive for counting applications, with a run capable of 24 focused human transcriptomes (via AmpliSeq) or 6 full human transcriptomes with a 12 hour sample-to-data run time (split evenly between prep and sequencing instrument time). This is claimed (I just haven't tried checking the math) to offer some of the lowest prices per read,  $3-$4 per million reads (with 24 hour runs on PII), on the market, enabling "precision of digital NGS cheaper than arrays".


A key addition to Ion's pantry is the HiQ chemistries for the two platforms is the HiQ polymerase, an engineered polymerase to greatly tamp down the indel error problem with the non-terminator Ion chemistry.  Ion believes that HiQ on Proton PI chips, which is now shipping, offers accuracy comparable to the Illumina NextSeq, though the indel issue has not been fully eliminated .

Ion Chef

To go beyond the soupรงon of information above, the Ion Chef is having its role expanded beyond just template preparation into other stages of the process; the vision is to have all pre-sequencer sample preparation occur on the $55K Chef.  For example, around mid-year AmpliSeq will be fully moved onto the Chef via a simple cartridge, enabling roughly 10 minutes of hands-on time to generate ready-to-sequence material from input DNA


Ion continues to keep their platform simmering in established niches, particularly in counting applications and where overall speed is key.  This is particularly important, given the ever-present spectre of new short read platforms in this space from QIAGEN, BioRad/GnuBIO and Roche/Genia, as well as the growing realization that Oxford Nanopore can deliver valuable data very fast.  Ion is building a platform focused greatly around the Ion Chef robot, expanding its generality within the platform, but also enabling rapid, single tube assays for specific clinical applications.  Ion may not have succeeded in conquering the sequencing world, but remains an important player.


Dale Yuzuki said...

Hi Keith, many thanks for publishing this.

It's so important that the word gets out that there are alternative choices. A few weeks ago at the NCI I heard a friend Stephen Chanock give the Jeffrey Trent Annual Award Presentation on the NIH campus, and when he pointed out that 'you know there are other platforms other than Illumina' I just about laughed aloud. There are many who just aren't paying attention to what is going on in the NGS landscape, and your piece can help enlighten.

Intereting nuance about Isothermal Amplification - the meso-reads will be so useful for HLA and TCR applications where targeted, accurate sequencing is so important. I thought I'd see much more from Pacific Biosciences on this one (may have missed it in the posters, alas I didn't have that luxury as in prior years due to other commitments... the 10' I had during the poster sessions I spent at 10x then to the Univ Fl poster on single-cell methylation.)

Hopefully we can catch up IRL at ASM this year. Thanks for your insight over Twitter too during #AGBT15 - the ONT hairpins being enriched by His-tags was quite the surprise!

All the best -

Raja Mugasimangalam said...

Hi Keith,
Thanks for the blogs - I am using your blogs to validate the rumors I get from around the world. I am also posting the weblink of your blogs at the world's largest Genomics -NGS/array linkedin group.
Keep writing