NovaSeq index issue (vs MiSeq)

JakobHedegaard · 12-19-2018, 02:57 AM

Hi,
We have conducted a NovaSeq S1 run of different library types and noticed a lower %perfect reads from a single sample as compared to a MiSeq QC run of the same libraries.
The libraries were a mixture of 21x QIAseq FX, 9x KAPA RNA Hyper libraries (both with 8 bp dual indexes) and 36x KAPA Hyper DNA libraries prepared with NextFlex48 single indexes from BiooScientific (cat#514104).
We notice that a single KAPA Hyper DNA library with index sequence GTAGAGAT+TCTTTCCC (Barcode Adapter 17) were found to have 76 % perfect index match on the NovaSeq while it had 97.8 % perfect index match on the MiSeq. The insert sequences from the NovaSeq were furthermore much more noisy from this specific library.
All other KAPA Hyper DNA libraries performed fine on both MiSeq and NovaSeq.
The NextFlex48 indexes are single 6 bp index but can be read as 8 bp dual indexes resulting in an additional and universal terminal AT and an universal TCTTTCCC index 2 from all KAPA Hyper DNA libraries (see the attached pdf).

Wonder why a single library can behave so different on a MiSeq and NovaSeq- while similar libraries essentially only differing in the index sequence behave identical.
I have asked the same question to Illumina Tech Support and was asked to pass the question on to BiooScientific. I have done so, but wonder if any of you experienced NovaSeq users have encountered similar issues with specific indexes.
Best, /Jakob

JakobHedegaard · 12-20-2018, 01:24 AM

Come on - I have heard several rumors about indexes behaving non-optimal on the NovaSeq. Please share your experiences :-)

GenoMax · 12-20-2018, 05:02 AM

MiSeq is a sequencing champ as far as illumina libraries are concerned. We look at results from MiSeq nano runs only qualitatively (rather than quantitatively) before the pools are transferred for final NovaSeq runs. You always need to keep in mind G = no signal limitation of 2-color chemistry as well.

If you are consistently observing that index combination performing badly then BioO would need to know that.

pmiguel · 12-20-2018, 12:40 PM

You are running non-UDI indexes on the NovaSeq? Are you not concerned about index swapping?

We demultiplex allowing 0 mismatches for all Illumina runs these days. So I haven't taken note of non-concordant perfect index matches between the MiSeq and NovaSeq.

However, this is an issue mentioned in the Illumina White Paper on using an iSeq to titrate reactions for a NovaSeq.
--
Phillip

luc · 12-29-2018, 01:39 PM

I would suggest to check the index reads for this particular library to see which base is affected.

JakobHedegaard · 01-15-2019, 05:03 AM

I have been told that two TruSeq UD indexes have shown lower than expected perfect index reads: UDI0025 (ACTAAGAT, AACCGCGG) and UDI0055 (AGGCAGAG, AGAATGCC).
Anyone else seen this?

JakobHedegaard · 01-15-2019, 05:12 AM

And in a recent S2 (2x150) NovaSeq run of eight Nextera DNA Flex libraries prepared from human gDNA using indexes from the CD Indexes (96 Indexes, 96 Samples, cat# 20018708) we observed a single library with only 24-27% perfect reads. The index was from well F04, H701 (TAAGGCGA) + H522 (TTATGCGA) and the top non-perfect index was TAAGGCGA+TTTTGCGA (so a T in pos 3 in index2 where it should have been a A).
The other indexes in the run were find with a high fraction of perfect indexes.
Other observations of indexes with high fraction of non-perfect indexes?
We allow a single mismatch so we do get the data we need. We are though a bit worried that the quality may be affected.
And it would be great with an explanation of this index issue on the NovaSeq!

Pichelsberg · 01-15-2019, 04:32 PM

Helpful thread.

JakobHedegaard · 01-18-2019, 02:18 AM

In our hands, we do not find any significant differences in the quality between insert data with perfect index reads and insert data with 1-mismatch index reads.
So despite a much larger fraction of 1-mismatch index reads the data is fine. Nice...

But it would be great with an explanation of how/why this index issue on the NovaSeq actually happen.
In our data, the issue is more frequent in index2 than in index1.
Can it have something to do with a more "loose" binding to the surface oligo when priming for reading index2? Or a sequence-specific higher (and systematic) error rate in index2 reading?
Any ideas?

And a list of indexes where one should expect a higher fraction of non-perfect indexes would be nice. Should we start one?

ralyma · 01-25-2019, 07:15 AM

Dear Jakob,

We also have observed problems with H522=TTATGCGA. In demultiplexed data (default settings accepting 1-mismatch) 97% of observed adapters are TTTTGCGA.

To my knowledge TTTTGCGA is not an Illumina adaptor, so this makes it less likely that we have mixed H522 with another adaptor. Then it is more likely that there has been has been an issue in the production of H522 so that sequence TTTTGCGA has been produced instead of TTATGCGA.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Novaseq vs Miseq failure	Tcom111	Illumina/Solexa	8	12-10-2018 04:01 AM
I7 index reading issue	johnathanlee	Illumina/Solexa	2	07-09-2018 01:27 AM
Miseq dual index Read1 (i7 index) problem?	Min Jung Lee	Illumina/Solexa	4	01-08-2018 12:22 PM
Pooling TruSeq (single index) and Nextera XT (dual index) in one MiSeq run?	Carosmile	Sample Prep / Library Generation	1	05-16-2017 02:13 PM

01-15-2019, 05:03 AM	#6
JakobHedegaard Member Location: Aarhus, Denmark Join Date: Mar 2008 Posts: 57	I have been told that two TruSeq UD indexes have shown lower than expected perfect index reads: UDI0025 (ACTAAGAT, AACCGCGG) and UDI0055 (AGGCAGAG, AGAATGCC). Anyone else seen this? Last edited by JakobHedegaard; 02-12-2019 at 11:52 AM.

12-20-2018, 01:24 AM	#2
JakobHedegaard Member Location: Aarhus, Denmark Join Date: Mar 2008 Posts: 57	Come on - I have heard several rumors about indexes behaving non-optimal on the NovaSeq. Please share your experiences :-)

12-20-2018, 05:02 AM	#3
GenoMax Senior Member Location: East Coast USA Join Date: Feb 2008 Posts: 7,082	MiSeq is a sequencing champ as far as illumina libraries are concerned. We look at results from MiSeq nano runs only qualitatively (rather than quantitatively) before the pools are transferred for final NovaSeq runs. You always need to keep in mind G = no signal limitation of 2-color chemistry as well. If you are consistently observing that index combination performing badly then BioO would need to know that.

12-20-2018, 12:40 PM	#4
pmiguel Senior Member Location: Purdue University, West Lafayette, Indiana Join Date: Aug 2008 Posts: 2,317	You are running non-UDI indexes on the NovaSeq? Are you not concerned about index swapping? We demultiplex allowing 0 mismatches for all Illumina runs these days. So I haven't taken note of non-concordant perfect index matches between the MiSeq and NovaSeq. However, this is an issue mentioned in the Illumina White Paper on using an iSeq to titrate reactions for a NovaSeq. -- Phillip

12-29-2018, 01:39 PM	#5
luc Senior Member Location: US Join Date: Dec 2010 Posts: 452	I would suggest to check the index reads for this particular library to see which base is affected.

01-15-2019, 05:12 AM	#7
JakobHedegaard Member Location: Aarhus, Denmark Join Date: Mar 2008 Posts: 57	And in a recent S2 (2x150) NovaSeq run of eight Nextera DNA Flex libraries prepared from human gDNA using indexes from the CD Indexes (96 Indexes, 96 Samples, cat# 20018708) we observed a single library with only 24-27% perfect reads. The index was from well F04, H701 (TAAGGCGA) + H522 (TTATGCGA) and the top non-perfect index was TAAGGCGA+TTTTGCGA (so a T in pos 3 in index2 where it should have been a A). The other indexes in the run were find with a high fraction of perfect indexes. Other observations of indexes with high fraction of non-perfect indexes? We allow a single mismatch so we do get the data we need. We are though a bit worried that the quality may be affected. And it would be great with an explanation of this index issue on the NovaSeq!

01-15-2019, 04:32 PM	#8
Pichelsberg Junior Member Location: new york Join Date: Dec 2018 Posts: 5	Helpful thread.

01-18-2019, 02:18 AM	#9
JakobHedegaard Member Location: Aarhus, Denmark Join Date: Mar 2008 Posts: 57	In our hands, we do not find any significant differences in the quality between insert data with perfect index reads and insert data with 1-mismatch index reads. So despite a much larger fraction of 1-mismatch index reads the data is fine. Nice... But it would be great with an explanation of how/why this index issue on the NovaSeq actually happen. In our data, the issue is more frequent in index2 than in index1. Can it have something to do with a more "loose" binding to the surface oligo when priming for reading index2? Or a sequence-specific higher (and systematic) error rate in index2 reading? Any ideas? And a list of indexes where one should expect a higher fraction of non-perfect indexes would be nice. Should we start one?

01-25-2019, 07:15 AM	#10
ralyma Junior Member Location: Copenhagen Join Date: Aug 2018 Posts: 1	Dear Jakob, We also have observed problems with H522=TTATGCGA. In demultiplexed data (default settings accepting 1-mismatch) 97% of observed adapters are TTTTGCGA. To my knowledge TTTTGCGA is not an Illumina adaptor, so this makes it less likely that we have mixed H522 with another adaptor. Then it is more likely that there has been has been an issue in the production of H522 so that sequence TTTTGCGA has been produced instead of TTATGCGA.