Tandem dual index- low Q30 for index 2

[juliar]

Hello,

While reading up about index hopping, I came across the Illumina whitepaper (https://www.illumina.com/content/dam...inkId=36607862) which states that unique dual indexes help to mitigate this.

While reading another paper about index hopping (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5759201/), where the authors used dual-matched indexes, they mention an online post by Illumina (https://support.illumina.com/bulleti...x-designs.html) which says that distinct dual indexes work better than tandem dual indexes. Illumina present data showing how tandem dual indexes have a low Q30 for the index 2 read in comparison with distinct dual indexes and data also which shows a ‘T overcall / A undercall’ phenotype for the tandem design. However, the authors of the dual-matched index paper did not find the reduced Q30 scores reported by Illumina.

So my question is why would using a tandem index affect the Q30 score for the second index read? I don't understand how data from the first index read would affect the data generated during the second index read?

Thanks in advance for any insight!

[luc]

The results do vary by sequencer (the Hiseq 4000 is not - or only minimally - bothered by dual matched indices); the NovaSeq OTOH can tank with these. Likely library insert sizes can also play a role.
The MiSeq and NextSeq also have no troubles with dual matched indices.

The clustering chemistry seems to be different on the NovaSeq but I have no idea what causes the problems.