Title: Scaling Semiconductor Sequencing.
- Sequencing on a semi conductor chip. Uses known technology for chips, but changes way we do sequencing. Chip is actually the machine.
- Should allow scaleup every 6 months.
- Chemistry is “post-light” sequencing. Chemistry requires no optics, so no quenching. Coverage is very uniform.
- Native polymerases and native nucleotides – which also makes it low cost.
- Chips start @ $250
- Speed is also a benefit. 2 hour seq. runs.
- Sample prep is decoupled from sample prep, which can be done in batch. Many improvements coming in sample prep – currently takes about 8 hours, should be about 2 hours by the time optimizations are done.
- Rapid performance improvements [nice graphic, but can’t copy it down for you… I’m sure they’d share it, however.]
Chad Nusbaum from Broad Institute will give the main talk:
Title: Implementation and applications…. something like that. [ (-: ]
- Key tech:
- Does it do something we need
- can it do it better
- can it provide special benefit?
- Is there enough need to make it worth the investment.
- What’s appealing about ion torrent:
- fast, simple, cheap
- “[Speed] matters more now than it did then.”
- Requires little infrastructure.
- “lightweight process”
- Easier to trouble shoot
- machine is small and inexpensive
- supply chain is simple too.
- Speed – quick turn around on processes.
- simplicity means esier optimization, etc
- Usual stuff
- Viral sequencing,
- QC large sequencing pools
- QC of targetted capture samples
- Exon capture experiments
- Genome assembly.
- Did a tumour sequencing validation expt.
- After filtering, you get 10-100 variations per tumour.
- showing validation results. 72/93 variants were called correctly (from the “hardest” set. [not sure why they were hard… microphone keeps cutting out, and may have missed something.]
- more than 99% of reads yield assignable barcodes (for a pooled library).
- Lab optimization: (slide on optimization cycle…. not particularly informative.)
- current chip have 7m wells.
- Observed low GC bias.
- good representation of homopolymers
- some computational approaches being worked on… wont talk about it.
Summary: fast, signifiant performances, demonstrated utility.
[Neat talk, and very quick.. now off to the meetup!]