Illumina just announced a new type of sequencer NovaSeq. Link for specifications.
NovaSeq 6000 - 6 Tb data/20 billion reads in 2 days (max possible output)
NovaSeq 5000 - 2 Tb data/1.6 billion reads in 2.5 days
- Four new flowcell types (S1,S2,S3,S4), 2 x 150 bp max, 2 x FC per sequencer.
- On board cluster generation.
- Two reagent cartridges. One for the cluster generation, the other for SBS reagents.
- New type of base call files (*.cbcl), can be converted to fastq.
- $850K for NovaSeq 5000 and $985K for NovaSeq 6000 (from Genohub blog)
A blog entry from Genohub indicates that these sequencers are using TWO color chemistry.
NOTE 1: Two color chemistry has been confirmed.
NOTE 2: Flowcell types can be mixed and matched. S1/S2 are restricted for NovaSeq 5000, all four can be used on NovaSeq 6000. Even though the flowcells have "lanes" like the NextSeq, they may have a single inlet port so only one sample pool per FC. Needs to be confirmed.
PS: If you were hoping to buy one be ready to wait. Apparently Illumina has pre-sold all NovaSeq they can make for a quarter or two this year.
Update 10/16/2017
S4 flowcells are now available. 48 human genomes per run (varbatim quote from Illumina). 6 Tb, 20 B reads, 2 x 150 bp.
There is also NovaSeq Xp workflow which allows loading of different pools in each lane. This would be essential for core facilities.
Update 01/24/17
Flowcells
S1,S2 - Flowcells for counting applications
For S2 - PF 1775-2070 cluster/mm^2
S3,S4 - High coverage applications
Same cartridges on both sides (three total).
Reagent Cartridge - Will take one library tube (200-300 pM) per FC.
SBS Reagents - Frozen
SBS Buffers - Room temperature ship/store. Ready to use.
- FC looks huge, like NextSeq. Easy to place. No futzing needed.
- Run setup identical to current HiSeq sequencers.
- Mix/match flowcells.
- Two sides of the instrument can work independently (stream data to BaseSpace on on side, store locally on other side).
- Automatic post-run wash once sequencing is complete
- Run times includes cluster generation.
- Network speed (200 mbps)
- ONE library pool per flowcell. Lanes visually distinct but fluidically same
- Both 5000/6000 are dual FC systems. For 5000 systems run times slightly longer
- Custom primers possible
- Asymmetric runs possible (e.g. 100 x 75)
- No "Rapid" run capability needed (Illumina's take)
- FC's can be used for single and paired-end recipes
- ExAMP, all self contained in the sequencer
- 200/240V, adequate HVAC
Test NovaSeq data is available on BaseSpace.
it can generate 132 exomes/transcriptomes per run
Automation, simple programming + scripting will be more important than ever
Focus with all new sequencing improvements seem singular. Sequence as many human samples as quickly as possible. Benefits derived by other disciplines are incidental.
Similarities between Illumina and Intel are amazing. Neither has much competition for the main application market they serve. Clever packaging with incremental upgrades ensures a continuous desire to upgrade on the part of consumers by ensuring products spread over the entire cost spectrum.
Without any information about reagent pricing it's hard to extrapolate to sequencing costs...
Also read length of 2x 0.15kb is disappointing. (Yes, read length is expressed in kb.)
But the removal of library/sample type restriction is a big win.
Ahh.... no. For the purposes of their specs, you have to sequence PhiX.
Think I'm kidding? Sad to say, that does not appear to be the case in practice (to claim failing specs on NextSeq, HiSeq 2500, or HiSeq1T). But at least you are allowed to sequence other things on this new platform, which is an improvement.
Yes, that's what I meant with
Can't help but note that the spec says nothing about the number of unique reads, or actual error rates rather than claimed quality scores...
Playing the devil's advocate -- when you have 10 Billion reads in the play, even if a full 10% are duplicates do the duplicates matter :)
Some interesting opinions and comments, with more information about the sequencers:
Blog post from Mick Watson
Blog post from Keith Robison
Blog post from James Hadfield
The guys at uni are questioning the assembly qualities coming out of the 2 colour chem vs the same library on a miseq with 4 colour still. There are supposed to be known biases in Base proportions. Anyone know more on this?
Yes, our 2-color chemistry assemblies only contain A and T, unfortunately; G and C are deprecated :(
Just kidding! Currently, we run HiSeq2500, MiSeq, and NextSeq. It's impossible to separate the difference between 2-color chemistry and other platform differences. NextSeq output is much lower quality than HiSeq2500 and MiSeq, so the assemblies are worse. And yes, there is high bias, particularly in read 2, where there is consistent base-frequency divergence:
But, that could be unrelated to the two-dye chemistry. It looks to me like it's more likely related to poor software calibration.
Please ignore the initial spikiness, which is an artifact of our fragmentation bias.
Our Illumina NovaSeq is on the way and ! expect to have the system up and running within a few weeks. Looking forward to S4 coming out. http://www.txgen.tamu.edu/