MiSeq - minimum index diversity?

[JBKri]

Hi, we want to do a MiSeq run of a library pool of TruSeq DNA PCR-free (single index; 6 bp), and custom 16s amplicons (dual index 8+8 bp). We wish the pool to contain 95 % TruSeq library and only 5 % of the amplicons. During the second index read especially, the sequence diversity will be very low if we pool like above. I calculated that the i5 index read will have only ~2 % diversity in every position. Is this enough to obtain good quality in the MiSeq? We could add more of the 16s amplicons to boost the diversity of the index reads. Should I aim for > 5% diversity? Adding more PhiX wouldn't help, since it has no index, right?
Jon

[bunce]

Hi Jon, It might be OK to run this with such low diversity - but I would not risk it. If it were me I would take a previous dual-indexed library (with different indexes to those in your library) and spike it in as well (~5%). It is always seemed crazy to me that PhiX is not indexed. I have thought about buying in some PhiX and building a nextera library with it so it can be indexed and easily identified (but should still automatically allign). cheers Mike

[Jessica_L]

If memory serves, PhiX used to have an index on it and a lot of end users got upset about the fact that one of the indices was always "taken". Granted, this was before there was dual indexing and you only had about 12 barcodes to choose from and a lot of users were running control lanes, not spike-ins; it's definitely less of an issue now that there's so many more barcodes available.

[JBKri]

Hi Jon, It might be OK to run this with such low diversity - but I would not risk it. If it were me I would take a previous dual-indexed library (with different indexes to those in your library) and spike it in as well (~5%). It is always seemed crazy to me that PhiX is not indexed. I have thought about buying in some PhiX and building a nextera library with it so it can be indexed and easily identified (but should still automatically allign). cheers Mike

Thanks, I thought about adding from a previous library, but I think we can manage by just increasing the proportion of the current 16s amplicons. By the way, there is indexed PhiX available (TailorMix), I might give it a try.
Jon

[Brian Bushnell]

Even if PhiX is indexed, that's still just one additional index. It's best to have multiple different bases at every position for barcode reading - they are low-quality enough even when everything is optimal. Potentially losing 40% of your data to the unknown bin due to miscalled barcodes is much worse than intentionally losing 10% of your lane to sequencing irrelevant stuff with other barcodes.

[JBKri]

So, we ended up doing the run with almost half and half of TruSeq library and 16s amplicons, which of course caused no problems in the index reads. But now we are again faced with this possible problem, since we want to do a run with 95 % TruSeq single index, and 5 % dual index amplicon.

Illumina has conflicting information regarding index color balance. In the latest "TruSeq Library Prep Pooling Guide" it says "It is important to maintain color balance for each base of the Index Read being sequenced, otherwise Index Read sequencing could fail due to registration failure" and "Check the color balance using IEM".
But IEM does not check color balance since v. 1.7. In the IEM user guide it says: "IEM*no longer checks the base pair diversity in the indexes specified for MiSeq Targeted Resequencing in the TruSeq
Amplicon workflow. MiSeq RTA v1.17.28 and higher processes low-plexity index reads for the TruSeq Amplicon workflow, which makes base pair diversity checks in IEM unnecessary".

IEM will in fact no longer give any warning if there is no index diversity at all. Still I am worried if a very low diversity could cause problems. Does anyone have experience with very low index diversity? Previously we did a run with 87 % of a single index, with excellent results.

[GenoMax]

As long as your TrueSeq single index is unique (choose it so that it is compatible with tag 1 of the dual indexed pool). For index read 2 you are going to get some garbage since there is no tag2 for the single index library. If you are at all worried then slightly underload the pool than what you do normally.

[JBKri]

As long as your TrueSeq single index is unique (choose it so that it is compatible with tag 1 of the dual indexed pool). For read 2 you are going to get some garbage since there is no tag2 for the single index library. If you are at all worried then slightly underload the pool than what you do normally.

thanks. We will probably do just that, underload slightly.
I assume you mean garbage for index 2, not the actual read2. For the TruSeq single index, the second index read should give part of the sequencing adapter. We have done this before, it demultiplexed just fine. We put the adapter sequence as index in the sample sheet (TCTTTCCC).

[GenoMax]

I assume you mean garbage for index 2, not the actual read2.

Ah yes. Corrected my post above.

[Jessica_L]

I've run a couple of MiSeq runs where there were few indices used (i.e. 2 or 3, incomplete color balance) and we didn't have any trouble demultiplexing. The newer versions of the MiSeq software, as you pointed out, are much better at doing this than they used to be.

[JBKri]

thanks. We will probably do just that, underload slightly.
I assume you mean garbage for index 2, not the actual read2. For the TruSeq single index, the second index read should give part of the sequencing adapter. We have done this before, it demultiplexed just fine. We put the adapter sequence as index in the sample sheet (TCTTTCCC).

Just to let you know, the run went just fine. The index reads looked completely normal even with several positions with ~97 % of the same base in he second index.