NextSeq500 early results

[HMorrison]

Hi all,

We replaced our Hiseq1000 with a NextSeq, largely due to the expense of the HS maintenance contract and the long run time. Our installation and training runs seemed fine, and I'm trying to assess how good our first 'real' run is. One thing I noticed is that reads are no longer of uniform length (44-152 nt on a 2x150 run). Another is that quality score binning is the default (which is okay for our purposes). I see quite a lot of N tails. We are planning to use it for 16s v6 sequencing, at a max of 1/6th of the flowcell combined with high diversity metagenomic or RNAseq libraries. Longer amplicons go on the Miseq, as before.

I'd be interested to hear early impressions from others who have made the switch and if there are any other tricks to moving Hiseq protocols to the Nextseq.

Thanks,
Hilary

[Brian Bushnell]

One thing I noticed is that reads are no longer of uniform length (44-152 nt on a 2x150 run).

It's not really possible for the raw reads to be variable length, so I suspect you have the software set to automatically do adapter trimming.

Another is that quality score binning is the default (which is okay for our purposes).

Rather than "default", it's forced and cannot be disabled.

There's a thread about NextSeq here:

http://seqanswers.com/forums/showthread.php?t=40741&highlight=nextseq

[HMorrison]

It's not really possible for the raw reads to be variable length, so I suspect you have the software set to automatically do adapter trimming.

Rather than "default", it's forced and cannot be disabled.

There's a thread about NextSeq here:
http://seqanswers.com/forums/showthr...hlight=nextseq

Thanks for the quick answers. I looked at the earlier thread, but didn't find much regarding quality. I think the adapter trimming crept in because of the new format of the SampleSheet.csv. The binning is less of a problem than the quality inflation.

[Brian Bushnell]

If quality-score inflation is a problem, you can recalibrate using CalcTrueQuality, which I wrote primarily for NextSeq reads. To create the calibration matrix, you need high diversity data, though - so just create it from the RNAseq or metagenomic reads, not from the amplicons. You can still recalibrate the amplicon reads using the matrix generated from the mapping of the other reads in the same library.

[rocksd]

Hi all,

We replaced our Hiseq1000 with a NextSeq, largely due to the expense of the HS maintenance contract and the long run time. Our installation and training runs seemed fine, and I'm trying to assess how good our first 'real' run is. One thing I noticed is that reads are no longer of uniform length (44-152 nt on a 2x150 run). Another is that quality score binning is the default (which is okay for our purposes). I see quite a lot of N tails. We are planning to use it for 16s v6 sequencing, at a max of 1/6th of the flowcell combined with high diversity metagenomic or RNAseq libraries. Longer amplicons go on the Miseq, as before.

I'd be interested to hear early impressions from others who have made the switch and if there are any other tricks to moving Hiseq protocols to the Nextseq.

Thanks,
Hilary

Hi HMorrison,

Are you satisfied with the quality of the NextSeq run? How does it compared to your HiSeq runs?

Kindly share your experience with us and thanks!

James

[HMorrison]

NextSeq500 early results

James,

I am very pleased with our early results. I spent a lot of time looking at quality scores and accuracy (based on phix, which was the only control we used). The only quality drop off prior to the last cycle was on 16s v6 tags (template ended before the run) and a metagenomic library where the size selection was sub-optimal (a high proportion of inserts shorter than 150 bp). I calculated the error rate of phix as .004 (errors/base) for Hiseq and 0.007 for Nextseq.

Hilary

[rocksd]

James,

I am very pleased with our early results. I spent a lot of time looking at quality scores and accuracy (based on phix, which was the only control we used). The only quality drop off prior to the last cycle was on 16s v6 tags (template ended before the run) and a metagenomic library where the size selection was sub-optimal (a high proportion of inserts shorter than 150 bp). I calculated the error rate of phix as .004 (errors/base) for Hiseq and 0.007 for Nextseq.

Hilary

Hi Hilary,

Thank you very much for the quick reply and the quality information! It is very helpful.

James