Ion Torrent Retrospective - 2011

We’ve come a long way in 6 short months.  The beginning of a new year is a great time for retrospectives.  Keith Robison over at Omics! Omics! has done just that with his “Reflecting on a Year of Ion Torrent”.  I want to do something similar, but will use this opportunity to revisit one of our first Ion posts and draw some comparisons.

I came across one of the original posts in our Ion series “Ion v. MiSeq – Is There a Competition, And If So, Why?”  I will stay away from the Miseq v. Ion debate in this post.  Rather, I would like to highlight the growth of the platform in throughput, quality, cost and usability.  I may even do a little bit of looking forward…though Keith has done sufficient prognostication - I would just be reiterating.


In May/June we had run a 314 chip of E. coli that clocked in at a whopping 24Mb.  At the time, Ion had just released some internal data they generated amounting to 175 Mb raw sequence data.  As you can see now, we are routinely generating over 250 Mb of data, “EdgeBio Posts 414.31MB Ion Torrent Run” But, you say, “Anyone can generate massive amounts of data…Is it correct?”  This run is 372.97 Mb of Q20 data and 335 Mb of PERFECT data when comparing back to a known reference.  With this increased throughput, and the introduction of barcode sets this summer, we continue to build out a flexible offering on the system for our clients.


6 to 9 months ago our quality histograms looked like this: 

Quality scores across all bases (Sanger / Illumina 1.9 encoding)

Now they look like this:

Quality scores across all bases (Sanger / Illumina 1.9 encoding)

The run is holding its quality longer and the high quality data is higher quality than before.  With higher quality, comes high accuracy in calling variants.  You can check out some detailed information of us picking apart the Ion and Miseq in our post “Variant Calling on Ion Torrent Data”.

Read Length

One can also see the evolution of the read length from these graphics.  From 100 to 200 to 300.  Our latest runs are looking even better.  Here is a Read Length histogram from a human sample we sequenced a couple days ago.

Read Length Histogram

In June we compared the ability of the Miseq and the Ion to perform de novo assembly of a microbial genome.  It was obvious that the MiSeq was producing a higher quality assembly compared to the Ion.  The N50 contig length was almost double (50K v. 98K), with a maximum contig size of 232K for Miseq and 211K for Ion.  Our latest run, still with no paired end data and still at 100bp, generated N50s of 85K and a largest contig size of 247K with an average consensus quality of 62.


6 months ago a run at EdgeBio of a 314 chip cost $2500 for ~25Mb of sequence.  That a buck for every 10,000 bases.  Now we charge $2350 for a 316 chip (the 314 is discounted to $1550) and generate on average 250Mb.  That’s a buck every 106,000 bases.  So, for the same price, we have doubled the assembly metrics in de Novo assemblies above. All done in less than 7-10 business days.  For more detailed service pricing, please see our Ion Torrent Website.

Extensibility and Usability

Everyone knows, or should, that it’s not reagents or machines that are expensive over the long haul - it's people.  Ion has introduced time saving mechanisms in both the lab and on the informatics side.  The One Touch has cut down our preps times, and the introduction of the Torrent Suite of Software right from the initial launch of the platform have saved us cold, hard cash.  With the open architecture of the Torrent Suite and Server, we have been able to quickly integrate plug-ins such as FastQC, SnpEff, MIRA, etc to quickly QC, analyze and assemble our data.  No other new machine in my 10 years of working in this field has so simply just worked – straight out of the box.

We have had our issues, and the platform isn’t perfect.  We understand the limitations of the system in terms of cost per base, homopolymer accuracy, etc – but understanding the limitation allows you to make smart decisions on how and when to use the platform or an alternative.  Something a service provider SHOULD be expert at.  I know that past performance doesn’t guarantee future success, but as we have seen the trajectory of the machine, we are confident that we will continue to use and push this machine as much as we have over the past 6 months – 300 runs to date…


We have recently been validating the long read chemistry further, have done our first 2 Ampliseq Cancer panel runs, are gearing up to validate the mate pair protocol, and are piloting the 318 Chips.  Look for a few blog posts over the coming weeks about the mate pair data, our custom SnpEff plug-in,  and our progress with capture and 318 chips (maybe Exomes you say???)

Very exciting time around Edge, and we look forward to what 2012 may bring.