Tuesday, October 22, 2013

Ion Previews More Accurate Polymerase, Faster Template Prep

I haven't talked about Ion Torrent for a while, because it was largely off my radar screen.  In early 2012 the PGM had been an important contributor to my early de novo genome assemblies, as it was the only fast turnaround, low cost system I could access.  But the data quality was always frustrating, with many indels, and the 200 basepair mode on the read lengths not great for assembly.  Once I could access a MiSeq, that became our dominant instrument for individual genome assembly.  We tried Ion once more with the 300 basepair chemistry, but were not particularly impressed.
Despite this, the Ion team reached out to me and invited me to come down to Ion World (conveniently across town) and chat frankly about their new HiQ chemistry.  While I had been napping, Ion had pushed the read lengths on the PGM to 400 basepairs, but indels still dogged them, as seen in various public comparisons of sequencing platforms.  The HiQ enzyme has two mutations engineered in which improve fidelity.  Systematic indel errors are reduced by 90%, to 0.027%, with strand biases in errors markedly reduced  They did confess to me that they have not yet tried the system in the G+C extreme zone I work in (>70%).

The team screened first point mutations targeted to a region of the polymerase they thought would be productive, and single mutations from that screen were then screened in combination to arrive at the double mutant.  The mutant is thought to open secondary structure more aggressively and maintain a more consistent speed of incorporation, hence generating a more nearly linear signal when working a homopolymer.

Something I had not realized is that Ion has bifurcated their chemistry.  While Jonathan Rothberg had always emphasized that Ion could march along a lot of progress in the semiconductor field, others (including myself) had wondered whether the biochemistry would trip things up.  Clearly, Ion's experience shows that silicon is easy, but enzymology is hard.  The PGM and Proton are using different polymerases, and the initial HiQ launch is only for the PGM.  Full launch isn't until mid-summer next year, but those wishing to try it out will be able to request a tube of enzyme through the Ion Community starting next month. 

Ion has experimented with a number of ways to reach out to their users in unusual ways, and the new program of enabling early technology access to anyone is certainly an interesting twist.  Most early access programs target a select few partners at large genome centers.  Sometimes one can get connected through professional networks (that's how Warp got a crack at Moleculo when they were in stealth mode). but other times one simply strikes out (for various reasons I won't list all the companies that have resisted my entreaties).  Ion takes a risk that early versions won't perform; they are betting that the early users will understand the preliminary nature of what they are getting and will find exciting new uses for them.

Ion also announced that their Avalanche template preparation technology will also be made available through the early access program.  Less happy news is that this tech will be apparently be marketed as Ion Isothermal Amplification Chemistry; perhaps their annual quota of naming magic was used up with "Ion Chef".  Avalanche replaces emulsion PCR with an isothermal, single-tube reaction in which clonal amplification is achieved in a gel; the gel matrix restricts the diffusion of product in the reaction.  In some ways this is "back to the future"; the original Church polony technology involved gels.  But, for IP reasons alone there must be differences, though Ion was coy about them.  Avalanche makes a huge difference, especially with longer inserts/reads.  With emPCR, 400 basepair inserts require eight hours of cycling; with Avalanche the entire process is 2 hours!

While I've not been investing in Ion technology, clearly they have a foothold in the clinical sequencing space.  I listed to several groups speak at Ion World, and have followed the tweet stream on others.  Boston Children's Hospital has spun out Claritas (which appears to be taking residence in Codon's final space), which just landed a contract from the Veterans Administration for a large exome sequencing project.  BGI also just placed a giant order of Protons.  Ion adopters often cite operating speed as an important consideration and cite SNP-accuracy comparable to other platforms; they've previously had to hold their nose on the indel issue.  Launching the new chemistries may also help Ion snag marooned 454 customers looking for a new home.

Clearly HiQ and Avalanche are addressing two key issues for the core constituency of the Ion platform.  Avalanche means fast and easy template preparation; no additional hardware and shortened times.  Of course, Life would be happy to take that savings and have you apply it to the Ion Chef, a specialized liquid handler which can prepare two chips per day from prepared libraries.  A key selling point of Ion Chef is that it loads the chips consistently; good loading is apparently a tricky art (and loading variability, via those colorful plots, could really be turned into art!).   

HiQ will address the accuracy concern, though initially only on PGM, so that will benefit in the clinical space only those running amplicon sets.  Given the popularity of exome sequencing on the Proton, Ion must be trying to replicate the same mutations in the polymerase used there (which they were quite confident can be done) and validate that chemistry.  

Ion particularly needs to get these improvements out because the improvements in the chips are coming slowly to Proton (as I said above, silicon is easy...).  The Proton PII chip, which was projected to produce 32Gb per run, is still stuck in the starting blocks months after its original release date.  When it breaks out, PII will support only 100 basepair reads versus the 200 bp  reads currently supported on PI.  The PIII chip, which would easily cover a human genome in a run at sufficient depth for variant calling, seems very far off given the challenges of continuing to miniaturize (which, apparently, even include electronics challenges with pulling so much data off the chip so rapidly!). 

I can't quite convince myself that my shop has a niche for Ion, but I'm glad they are making some important improvements.  Illumina needs some real competition to spur further development, and only continued nipping at their heels by Ion and PacBio will supply that until serious new entrants (QIAGEN? GnuBio?) appear.  Even if you don't plan to use Ion sequencers, you really should be applauding their efforts.

4 comments:

Anonymous said...

Any thoughts on ZSGenetics and using an EM to do sequencing? I'd think with their claims of extra long reads you'd be all over that like a fly on...well you know. :)

Keith Robison said...

ZS is worth watching; I haven't seen a presentation in a while. Among the questions around their technology is distinguishing 4 bases, reliably labeling long stretches, stretching the DNA & how expensive their box will be. Still, super long reads are possible.

Dale Yuzuki said...

Hi Keith, great writeup. (Now I realize a number of blog posts I missed during all the festivities of last week's Ion World and ASHG meetings.)

One minor clarification - the OneTouch 2 / OneTouch ES now takes about 5.5h with the 400bp chemistry, not 8h as it was several iterations back.

I'm sure the naming group will come up with a catchy phrase to replace the term 'Avalanche'. But in my own experience, naming things sure is difficult.

Lastly, regarding ZS, this local MA news item announces they have expanded into a new 10K sf facility, and are up to 10 employees now.

All the best - Dale

bping said...

Hi really good review of the current state of play of the PGM. One simple question:

If both platforms were the same price (Miseq and PGM) and you wanted to do targeted resequencing for cancer diagnostics using their cancer panels and maybe develop your own panels down the road, which platform would you pick and why?

Any advice would be extremely useful as you have experience with both!

thanks