HiSeq 4000 index swaps

[GenoMax]

I was recently made aware of a potential problem that supposedly can result in index misassignments with HiSeq 4000 (under certain experimental conditions). James Hadfield wrote about this a few months back. This may be generally applicable for all ExAmp chemistry instruments.

Since I have not seen this discussed here/elsewhere online I wanted to see if people have seen this in your data.

[SNPsaurus]

Here's a preprint about it: http://www.biorxiv.org/content/early/2017/04/09/125724

[gringer]

There's also this thread.

[GenoMax]

Originally Posted by gringer

There's also this thread.

Thanks for the reminder @ginger.

[ECO]

Anyone doing significant multiplexing should be using dual indicies, which reduces this problem dramatically (making swapped reads go to Undetermined while demultiplexing).

Interested to see if this is a problem in our novaseq runs...will update if I note something.

[gringer]

That's dual indexes using the same index at both ends.

[DNATECH]

It is great that the Biorxiv manuscript from the Weissman lab has hopefully identified the cause of the problem which has been reported previously for some cases (the agent being "free barcoded primers" present in the library).
However, the scary index-swapping rates were seemingly generated from low-quality library preps with high primer contamination. I am aware that for their specific application (single-cell SMRT-seq) one can't be choosy with regards to the library quality. For any other application these libraries should have failed the QC. The free primers could and should have been easily drastically reduced by an additional magnetic bead cleanup for their first experiment. It is nice that we get to benefit from this mishap now.

I have posted my thoughts here: http://dnatech.genomecenter.ucdavis....ignment-issue/

Obviously these data ask for caution for any multiplexed sequencing projects and for protocol adjustments.
To some degree the manuscript shows: Ugly things can happen if one sequences really ugly libraries.
How relevant is this to the sequencing of high-quality and clean libraries? The observed linear correlation between between primer spike-ins and artifact rate indicates, to me, that there is no reason to panic.

Lutz

[SNPsaurus]

Good post, DNATECH. People are definitely overstating things on twitter, "10% reads misassigned on HS4000" without noting the caveats--this type of swap only happens when the primer contains indices, the paper used highly multiplexed samples where the read count per sample was low and the number of other samples that could contribute to crosstalk was high, and the prep (as you noted) lended itself to this issue.

I think it is also important to note that other forms of index swapping can occur (PCR of pooled libraries, for example).

[nucacidhunter]

Treating final PCR reactions (for library preps using Nextera, NEBNext line of kits, 10X Genomics and some other mentioned in DNATECH blog) with ExoSAP-IT or ssDNA nucleases such as Exonuclease1 should digest remaining PCR primers and can be adapted as an add-on before bead clean up.

Edit: Exonuclease VII also can be used to cleave primers and single stranded region of adapters.

[DNATECH]

nucacidhunter, this is definitely worth a try! (Is seemingly clean - according to BA - good enough?)

I agree SNPsaurus the problem is more complex.

Originally Posted by nucacidhunter

Treating final PCR reactions (for library preps using Nextera, NEBNext line of kits, 10X Genomics and some other mentioned in DNATECH blog) with ExoSAP-IT or ssDNA nucleases such as Exonuclease1 should digest remaining PCR primers and can be adapted as an add-on before bead clean up.

[Olaf Blue]

Treatment of PCR products with Exo I/EXOSAP/Antarctic Phosphatase for primer removal is an old trick and works very well.

[nucacidhunter]

Originally Posted by DNATECH

(Is seemingly clean - according to BA - good enough?)

It depends on:
1- Primer length that should be long enough not to blend with lower marker of the Chip or Tape. A 50 base long primer would blend with HSD5000 Tape lower marker but not with HSD1000 one
2- Sensitivity of Chip or Tape that should be high sensitivity type capable of detecting low concentrations

It should also be possible to Exo treat the library pool rather than treating each individual library

[nucacidhunter]

Illumina’s white paper on index misassignment:

https://www.illumina.com/content/dam...inkId=36607862

[GenoMax]

Originally Posted by nucacidhunter

Illumina’s white paper on index misassignment:

https://www.illumina.com/content/dam...inkId=36607862

Thank you @nucacidhunter.

[gringer]

1.5% index hopping with no adapters spiked in. Does that mean anything?

[nucacidhunter]

Originally Posted by gringer

1.5% index hopping with no adapters spiked in. Does that mean anything?

I think 1.5% hopping is from the libraries in the pool and not the spiked-in adapters.

[austinso]

Makes me wonder if single molecule consensus is out of the question, since it would seem that the mitigation strategy requires defined barcode sequences.

[DNATECH]

It shows that PCR-free libraries can hide a larger amount of adapters and adapter dimers somehow entangled in the y-adapters. We have several times seen surprises with (on the BA) clean looking PCR-free libraries in this regard.

Originally Posted by gringer

1.5% index hopping with no adapters spiked in. Does that mean anything?

[nucacidhunter]

Originally Posted by ECO

Anyone doing significant multiplexing should be using dual indicies, which reduces this problem dramatically (making swapped reads go to Undetermined while demultiplexing).

Interested to see if this is a problem in our novaseq runs...will update if I note something.

I wonder if there is any update on this.

[kmcarr]

Originally Posted by nucacidhunter

I wonder if there is any update on this.

In our core we did the following test: Project with 64 RNA-Seq libraries prepared using Illumina TruSeq Stranded mRNA HT Dual Index kit on a Perkin Elmer Sciclone G3 robot. We also add a second AMPureXP cleanup at the end of the prep. We divided these into 8 pools of 8 libraries each following the dual index pooling guidelines in Illumina's white paper about index hoping so that each pool had 8 dual-unique index pairs. Each pool was loaded on one lane of HiSeq 4K flow cell and sequenced SE50 (51 cycles R1, 8 cycles I1, 8 cycles I2) using standard protocol.

Total output per lane was ~370-380M reads. With 8 i7 and 8 i5 indexes in each lane there are a total of 64 combinations possible, 8 expected combinations and 56 which may result from index hoping. I ran bcl2fastq (2.19.0) looking for all index pairs, permitting 1 mismatch per barcode.

For any given unexpected index pair produced through index hoping I found an average of ~35,000 (± 9,850) reads or approximately 0.01% of the total lane output.