HiSeq 4000 and NovaSeq Sequencing Concerns
As you may have heard around campus, there has been some discussion on the integrity of multiplexed samples on Illumina HiSeq 4000 machines and those with similar chemistry and flow cells, such as the NovaSeq. If free barcoded adapter / index primers are present in a multiplexed pool, the free adapter has the potential to prime and extend library molecules in the same lane during the clustering step. This can result in mis-assignment of reads through index swapping. This can cause errors in demultiplexing data, as reads from one sample have the potential to end up in the FASTQ files of a different sample. The HiSeq 2000/2500 and MiSeq are less impacted due to their biochemistry and the geometry of the flow cell used.
The range of mis-assignment can vary significantly and is impacted by the following factors:
- Amount of free adapter present in library
- Storage conditions of library
- Application or library prep workflow
The exact percentage of reads affected varies, it is less than 0.5%. It is likely that the amount of read mis-assignment through index swapping is proportional to the amount of free adapter or index primer present in the multiplexed library. While this issue can affect all multiplexed samples, it has the highest impact on low diversity and/or low input libraries such as Single Cell, Cell Free DNA, or FFPE. The low amount of starting input may leave an excess of adapters or index primers following library preparation if no adjustment of insert to oligo ratios is made. Samples identifying low-frequency variants or highly noise-sensitive studies should be confirmed by use of orthogonal methods, while RNA-Seq and similar counting studies will benefit from pooling of similar samples and through the use of unique dual indexing strategies as outlined below.
This issue has only been recently discovered in our lab and we have also brought it to Illumina’s attention. For additional details, please see James Hadfield’s technical take on the subject in his article: Index mis-assignment between samples on HiSeq 4000 and X-Ten.
Sample mis-assignment can potentially impact users depending on the experimental design and library prep workflow. Illumina has been working on this issue internally and has developed a few suggested mitigation strategies to reduce index swaps, listed below:
During Library Construction:
- Optimize your PCR or ligation step to avoid an excess of adapters or index primers.
- For PCR dilute the index primers to adjust the insert to adapter / primer ratio.
- Perform extra clean ups after this step.
- PAGE purification seems to do a good job reducing indexing primers.
- Purification columns are also an option.
- Do extra clean ups of each individual library before pooling.
- Use single use aliquoted adapters and primers.
- Freeze individual libraries and pool prior to sequencing.
Pooling suggestions:
- Use dual indexing strategies with unique barcodes on both ends. (Swapping would have to occur at both ends for read mis-assignment to occur)
- Sequence or freeze created libraries pools as soon as possible.
Sequencing suggestions:
- Use PhiX from third parties with unique indexing barcodes to determine swap frequency. (We will have begun to introduce PhiX with unique barcodes from SeqMatic for HiSeq 4000 runs.)
- For methods highly sensitive to mis-assignment use HiSeq 2000/2500 or MiSeq instruments.
Ilumina also suggests that you follow Illumina Best Practices.
If you have questions about your results or samples, please contact Illumina Technical Support for additional help. They and the local Field Applications Scientists will be happy to assist you.
Email: TechSupport@illumina.com
Phone: 800.809.4566