MiSeq v3 600 bp improvements?

jhi_pete · 01-30-2017, 07:59 AM

I have heard a rumour that the long standing MiSeq v3 600 bp kit chemistry problems which led to poor quality read 2 data have recently been solved by Illumina. Anyone know if this is true?

nucacidhunter · 01-30-2017, 08:34 AM

It is true and for high diversity libraries output is according to specification. Amplicon sequence quality also has improved but as before cluster density should be reduced and spiked in with PhiX.

microgirl123 · 01-30-2017, 10:55 AM

I just started my first 600-cycle run in over a year so hopefully it's true! I'll try to update in a couple of days when it's finished.

plyguy69 · 01-31-2017, 08:31 AM

Can anyone confirm this? I need to do some 2x300 sequencing for Ig RNA-seq experiments. Would be awesome if I could use the MiSeq.
Thanks,

Markiyan · 01-31-2017, 08:32 AM

Yes, the last few runs were way better (touch the wood).

If doing amplicons, please use a lot of phiX or shotgun library (up to 20%) (esp. if fist 25 bp's are the same). It would not hurt, and gains in data quality/yield from better dephasing calculation far outweigh the losses from 5%-20% phiX sequence spike in.

If doing cDNA amplicon analysis: make sure you do not use oligoC oligos in your amplicons, because the read quality would go through the floor if there are 10-14C's or G's in a row, and you may also get non-specific amplification of rRNA (rcDNA), reducing usable data yield by 1-2 orders of magnitude.

Also lower the loading density for 500-600bp amplicons, if you need high quality reads for your analysis.

Obviously do not forget, that the longer the amplicon or the read is, the more sensitive it becomes to the DNA sequence content.

Brian Bushnell · 01-31-2017, 06:18 PM

Quote:

Originally Posted by plyguy69

Can anyone confirm this? I need to do some 2x300 sequencing for Ig RNA-seq experiments. Would be awesome if I could use the MiSeq.

Are there alternatives other than Sanger? Or, what was your backup plan?

@Markiyan - thanks for the advice, I'll pass it on!

jhi_pete · 02-01-2017, 01:20 AM

Thanks to everyone for their positive replies, that is very good news. We have ordered up a 600 bp kit and will test it ourselves soon and let you know how we get on.

Markiyan · 02-01-2017, 02:09 AM

Quote:

Originally Posted by Brian Bushnell

Are there alternatives other than Sanger? Or, what was your backup plan?

1. Pacbio RSII CCS - should work quite nicely for amplicons up to 2-4kb.

2. If you can cope with 2-4% systematic errors - Oxford Nanopore 2D.

3. There used to be Roche's systems - FLX+ and Junior, but the reagents for them are no longer available... :-(

jhi_pete · 02-21-2017, 05:49 AM

Quote:

Originally Posted by microgirl123

I just started my first 600-cycle run in over a year so hopefully it's true! I'll try to update in a couple of days when it's finished.

Hi Microgirl. How was your 600 cycle run?

microgirl123 · 02-21-2017, 05:53 AM

Not good

It was an amplicon run, but the cluster density was good (~600K) with 20% phiX, so I expected the Q30 to be much better than 64% overall. It was only 54% for Read 2. Both reads only got out to about 150 bp before dropping off dramatically. The investigator decided it was definitely not worth trying to use the 600-cycle kit again!

jhi_pete · 02-21-2017, 05:55 AM

Quote:

Originally Posted by microgirl123

Not good

It was an amplicon run, but the cluster density was good (~600K) with 20% phiX, so I expected the Q30 to be much better than 64% overall. It was only 54% for Read 2. Both reads only got out to about 150 bp before dropping off dramatically. The investigator decided it was definitely not worth trying to use the 600-cycle kit again!

Oh dear! That's unfortunate, sorry to hear that.

Anyone else with recent experience running the 'new' 600 bp kits? We plan to use one this week or next....

GenoMax · 02-21-2017, 05:59 AM

Quote:

Originally Posted by microgirl123

Not good

It was an amplicon run, but the cluster density was good (~600K) with 20% phiX, so I expected the Q30 to be much better than 64% overall. It was only 54% for Read 2. Both reads only got out to about 150 bp before dropping off dramatically. The investigator decided it was definitely not worth trying to use the 600-cycle kit again!

Am curious about the focus on the Q-scores. Is the investigator looking to call SNPs? Aligning to reference sequence should be no problem irrespective what the Q-score is.

Markiyan · 02-21-2017, 07:39 AM

Quote:

Originally Posted by microgirl123

Not good

It was an amplicon run, but the cluster density was good (~600K) with 20% phiX, so I expected the Q30 to be much better than 64% overall. It was only 54% for Read 2. Both reads only got out to about 150 bp before dropping off dramatically.

This looks like issues with adapter dimers to me - what was yours beads ampure cleanup ratios, and any slight bump up on the agilent trace in the region of 100-200bp?

Basically significant amount of the clusters on the run contained shorter templates, and once the end of these is reached (140-160 bp), it wrecks RTA dephasing calculations.

since templates in the 150-200 bp cluster very efficiently compared to 500-800 bp ones, it takes only 3-5% of the shorter contaminant to wreck Q scores in the run.

To minimize it avoid repetitive freeze thaw cycles of the Illumina adapters (they can lose T tails and self ligate) and make sure no nucleases contamination is present during the ligation.

thermophile · 02-21-2017, 10:29 AM

Quote:

Originally Posted by microgirl123

Not good

It was an amplicon run, but the cluster density was good (~600K) with 20% phiX, so I expected the Q30 to be much better than 64% overall. It was only 54% for Read 2. Both reads only got out to about 150 bp before dropping off dramatically. The investigator decided it was definitely not worth trying to use the 600-cycle kit again!

Tech support has told me not to cluster below ~750k even for v2 even for amplicon runs. If it's 16S, I had a particular set of samples from some sort of bird guts that I basically can't run by themselves even with 20% phiX because R2 is so bad. There seems to be something happening between cycles 10-30 of R2 that is actually sequence dependent. The way my facility is set up, I mix and match small projects on runs so this is addressable by splitting up the bird samples across many runs.

The suggestion that there may be a primer diamer issue is reasonable, but you also may need to up your density a bit.

nucacidhunter · 02-21-2017, 06:36 PM

Quote:

Originally Posted by microgirl123

Not good

It was an amplicon run, but the cluster density was good (~600K) with 20% phiX, so I expected the Q30 to be much better than 64% overall. It was only 54% for Read 2. Both reads only got out to about 150 bp before dropping off dramatically. The investigator decided it was definitely not worth trying to use the 600-cycle kit again!

In addition to above comments it could be a bad batch as well. My current runs Q30 are above %70 specified by Illumina for 2x300 reads:

fanli · 02-22-2017, 08:03 AM

We did a 2x300 run for a TCR library recently. Passed spec (77.5% Q30 @ 939k with 15% PhiX), but clear tailing towards the ends of both R1 and R2 as well.

I could see error rates approaching 5% being problematic for many if not most applications.

We'll be sticking with 2x250 or 2x150 chemistry for the time being, unless specifically requested by a client.

nucacidhunter · 02-23-2017, 02:43 PM

To my knowledge error rate is not included in Illumina specifications. Q scores drop after cycle 250 is a known characteristic of the chemistry and for some applications it will not be detrimental.

Bukowski · 03-23-2017, 02:50 AM

So.. no this issue is not resolved. I have an email from Illumina today which I will summarise.

The issue is that there is a byproduct in their manufacturing process of fluorescent nucleotides. This byproduct results in NON-CLEAVABLE NTPs - so they limit strand growth. Obviously this has limited effects on the number of clusters per cycle, but does become more acute on 600 cycle kits as the problem 'builds'. You get a lower cluster density, and lower Q30 and higher error rates on the later cycles.

Illumina can now *detect* this byproduct and know what levels impact performance, which sounds to me like there might be batch to batch variation, and have put in place measures to screen for it, therefore they are still expecting improvements 'in the near future' as newer material hits the manufacture pipeline. I am also told there will be a new announcement soon, with new timelines for resolution.

plyguy69 · 03-23-2017, 06:27 AM

I've have also heard that the problem wasn't resolved, but never had details as to the cause.

Thanks!

jhi_pete · 03-23-2017, 06:47 AM

Many thanks Bukowski, that's the first logical explanation I have heard. Looking forward to any announcements from Illumina going forward.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
low Q30 in index, miseq v3 600	cnano	Illumina/Solexa	3	06-05-2019 01:43 PM
Miseq v3 600 cycle kit - poor read 1 quality	incognitos	Illumina/Solexa	3	11-09-2016 10:46 AM
Problems with kit 600 cycles	Mario Moreno	Illumina/Solexa	14	04-22-2015 04:12 PM
MiSeq error rate for 600 cycle PE	Luvs_2_sequence	Illumina/Solexa	27	09-29-2014 12:53 PM
Performance improvements	perencia	Bioinformatics	5	07-02-2010 07:08 AM

01-30-2017, 07:59 AM	#1
jhi_pete Junior Member Location: Scotland Join Date: Jan 2013 Posts: 8	MiSeq v3 600 bp improvements? I have heard a rumour that the long standing MiSeq v3 600 bp kit chemistry problems which led to poor quality read 2 data have recently been solved by Illumina. Anyone know if this is true?

01-31-2017, 08:32 AM	#5
Markiyan Senior Member Location: Cambridge Join Date: Sep 2010 Posts: 116	Yes, the last few 2x300 runs were way better than 1 year ago. Yes, the last few runs were way better (touch the wood). If doing amplicons, please use a lot of phiX or shotgun library (up to 20%) (esp. if fist 25 bp's are the same). It would not hurt, and gains in data quality/yield from better dephasing calculation far outweigh the losses from 5%-20% phiX sequence spike in. If doing cDNA amplicon analysis: make sure you do not use oligoC oligos in your amplicons, because the read quality would go through the floor if there are 10-14C's or G's in a row, and you may also get non-specific amplification of rRNA (rcDNA), reducing usable data yield by 1-2 orders of magnitude. Also lower the loading density for 500-600bp amplicons, if you need high quality reads for your analysis. Obviously do not forget, that the longer the amplicon or the read is, the more sensitive it becomes to the DNA sequence content.

03-23-2017, 06:27 AM	#19
plyguy69 Junior Member Location: america Join Date: Apr 2015 Posts: 9	thanks for the update I've have also heard that the problem wasn't resolved, but never had details as to the cause. Thanks!

01-30-2017, 08:34 AM	#2
nucacidhunter Jafar Jabbari Location: Melbourne Join Date: Jan 2013 Posts: 1,238	It is true and for high diversity libraries output is according to specification. Amplicon sequence quality also has improved but as before cluster density should be reduced and spiked in with PhiX.

01-30-2017, 10:55 AM	#3
microgirl123 Senior Member Location: New England Join Date: Jun 2012 Posts: 200	I just started my first 600-cycle run in over a year so hopefully it's true! I'll try to update in a couple of days when it's finished.

01-31-2017, 08:31 AM	#4
plyguy69 Junior Member Location: america Join Date: Apr 2015 Posts: 9	Can anyone confirm this? I need to do some 2x300 sequencing for Ig RNA-seq experiments. Would be awesome if I could use the MiSeq. Thanks,

02-01-2017, 01:20 AM	#7
jhi_pete Junior Member Location: Scotland Join Date: Jan 2013 Posts: 8	Thanks to everyone for their positive replies, that is very good news. We have ordered up a 600 bp kit and will test it ourselves soon and let you know how we get on.

02-21-2017, 05:53 AM	#10
microgirl123 Senior Member Location: New England Join Date: Jun 2012 Posts: 200	Not good It was an amplicon run, but the cluster density was good (~600K) with 20% phiX, so I expected the Q30 to be much better than 64% overall. It was only 54% for Read 2. Both reads only got out to about 150 bp before dropping off dramatically. The investigator decided it was definitely not worth trying to use the 600-cycle kit again!

02-23-2017, 02:43 PM	#17
nucacidhunter Jafar Jabbari Location: Melbourne Join Date: Jan 2013 Posts: 1,238	To my knowledge error rate is not included in Illumina specifications. Q scores drop after cycle 250 is a known characteristic of the chemistry and for some applications it will not be detrimental.

03-23-2017, 02:50 AM	#18
Bukowski Senior Member Location: UK Join Date: Jan 2010 Posts: 390	So.. no this issue is not resolved. I have an email from Illumina today which I will summarise. The issue is that there is a byproduct in their manufacturing process of fluorescent nucleotides. This byproduct results in NON-CLEAVABLE NTPs - so they limit strand growth. Obviously this has limited effects on the number of clusters per cycle, but does become more acute on 600 cycle kits as the problem 'builds'. You get a lower cluster density, and lower Q30 and higher error rates on the later cycles. Illumina can now detect this byproduct and know what levels impact performance, which sounds to me like there might be batch to batch variation, and have put in place measures to screen for it, therefore they are still expecting improvements 'in the near future' as newer material hits the manufacture pipeline. I am also told there will be a new announcement soon, with new timelines for resolution.

03-23-2017, 06:47 AM	#20
jhi_pete Junior Member Location: Scotland Join Date: Jan 2013 Posts: 8	Many thanks Bukowski, that's the first logical explanation I have heard. Looking forward to any announcements from Illumina going forward.