TruSeq index mixing strategies

[captainentropy]

I can't find this answer anywhere. I was told that there are some combinations of barcodes/indexes that should be avoided in a lane.

Does anyone have a list of the least problematic combinations?

[pmiguel]

Hi captainentropy,

Well, there should be! Maybe there was at one time? We have only been doing Illumina runs for a few months. For the most part Illumina seems to have their act together in ways Roche and Applied Biosystems do not. However this is one case where Roche and ABI seem to have worked out issues that Illumina is letting their users stumble over.

The only guidance in this matter I have been able to find came in another thread from ETHANol. He recommended this link:

http://www.chem.agilent.com/Library/...exed_v.1.2.pdf

I quote the relevant passage from "SureSelect XT Target Enrichment System Kit for Illumina Multiplexed Sequencing" pp 54-55 below:

Some sequencing experiments require the use of fewer than 12 index sequences in a lane with a high cluster density. In such cases, select indexes carefully to ensure optimum base calling and demultiplexing by having different bases at each cycle of the index read. Illumina recommends the following sets of indexes for low-level pooling experiments.
Pool of 2 samples:
• Index #6 GCCAAT
• Index #12 CTTGTA
Pool of 3 samples:
• Index #4 TGACCA
• Index #6 GCCAAT
• Index #12 CTTGTA
Pool of 6 samples:
• Index #2 CGATGT
• Index #4 TGACCA
• Index #5 ACAGTG
• Index #6 GCCAAT
• Index #7 CAGATC
• Index #12 CTTGTA

I think the main idea is that there be clusters visible during each scanning pass so the instrument doesn't decide it has fallen out of focus and attempt to focus on a blank lane. If I understand correctly our sequencer (HiScanSQ) does A/C and G/T in separate passes. So if at least, say, 10% of the clusters are lit during both scanning passes, the instrument won't get confused. That is, any mixture of indexes that includes both an "M" (A or C) and a "K" (G or T) at each of its six positions should be okay. For the HiScanSQ. For a HiSeq, you would want to check to make sure that it scans A/C and G/T in those pairs. Probably, I guess. Or do mixtures that have all 4 bases at each position of the index.

--
Phillip

[HESmith]

As far as avoiding particular combinations, the TruSeq adapters are designed so that a single error in the index sequence read does not convert it to another index sequence (thereby preventing assignment of the read to the wrong sample). CASAVA allows the option of tolerating zero or one mismatches in the index during demultiplexing.

-Harold

[epistatic]

The information that Agilent posts comes from the Illumina Multiplexing sample preparation guide. The latest version I have (rev. D, 2/2011) has on page 39 the same "low-level pooling" recommendations.

Indexes:
The twelve supplied sequences are unique and can be distinguished from one another by considering only the first three nucleotides. In addition, no sequence has more than three positions of agreement with any of the others. In this way they have been engineered to be “self‐correcting” in that an index with a single error will still match better to the correct index than to any one of the other indexes, and can therefore still be used. The twelve indexes are as follows:

Low-Level Pooling:
Some sequencing experiments require the use of fewer than 12 index sequences in a lane with a high cluster density. In such cases, a careful selection of indexes is required to ensure optimum cluster discrimination by having different bases at each cycle of the index read. Illumina recommends the following sets of indexes for low‐level pooling experiments.

2 samples: 6, 12; 3 samples: 4,6,12; 6 samples: 2,4,5,6,7,12.

The phix control lane contains single index (#3).

[ashish]

The phix control lane contains single index (#3).

Only if you are using PhiX v2. The new v3 is unindexed.

[pmiguel]

Only if you are using PhiX v2. The new v3 is unindexed.

Yes, isn't that odd? The idea is that the unindexed PhiX just ends up in the index-unassigned "bilge" along with low-quality clusters that failed to demultiplex. But we sometimes include a non-indexed library mixed in with indexed samples. So I don't really like that idea.

You can still get indexed PhiX (v2), just buy that weird multiplex kit:

PE-400-2002 | Multiplexing Sequencing Primer and PhiX Control Kit v2.0 (10 flow cells)

--
Phillip

[gringer]

Also see this thread about indexes influencing spot identification:

http://seqanswers.com/forums/showthread.php?t=14313