Mixing different types of indexes on the same MiSeq run ??

[memento]

Is it possible to mix lets say, two libraries in the same run, but the indexing would be 7nt for first one, and 8nt for the second one ? Of course the shorter one is completely different (index sequences) from the longer one. Also would it possible to mix the above scenario assuming that one library has i5 and i7 indexing, while the other one has only i5 indexing ? How the software would interpret it ?

[Genohub]

Mixing indices

The demultiplexer in CASAVA works by parsing the sequence data from the designated index cycles. The main caveats are that the index read length has to be static, meaning that you cannot have your custom indexes of different lengths. The other caveat is that the index sequences are parsed out exactly unless you change from the default setting.

Looking at your question, it may be possible to demultiplex your sample. If you truncate all the indices from 8 to 7 bases and you still have a unique combination you could proceed that way to resolve your situation. You may not be able to allow for mismatches while doing this but you should be able to give it a try. If there is a duplication in the collection of indices, the software will not allow the demultiplexing to start and may give you an error. Checking your sample sheet to try and avoid this would be advised.

Another option is to run Casava analysis twice, first the whole flowcell with your 8 indexes and afterwards again with 7. Assuming these barcodes are in different lanes, you would discard the lane information which don’t have 8bases for the 8 bases analysis and the same for the second analysis.
In case you need to completely reanalyze your whole data set, cif files would be necessary to analyze them with the off line base caller. I don’t think that will be necessary in your case.

-Genohub

[memento]

thats a great point, truncating the index from 8 to 7 (as long as the indices are "different" enough)...
thanks!

[mastal]

Mixing different types of indexes on the same MiSeq run ??

Assuming these barcodes are in different lanes, you would discard the lane information which don’t have 8bases for the 8 bases analysis and the same for the second analysis.

-Genohub

As far as I know, the MiSeq has only one lane.

[Genohub]

Right, the MiSeq only has one lane. I didn't see the title of the post, so I had HiSeq in mind. The truncation will still work, even with one lane.

[Calluna]

I haven't tried it for mixed length indices, but I have done it for a run of dual and single index samples. In that case, you can make the sample sheet for a dual index run and use the adapter sequences for the i5 sequence (TCTTTCCC for Truseq PE, reverse complement of that if running a SR). See this thread: http://seqanswers.com/forums/showthread.php?t=26547

[mcnelson.phd]

We mix samples with different types of indices all the time. In fact, we have a run going right now that has 16S samples with 12 base indices, Nextera XT libraries with 8 base dual indexing, and TruSeq libraries that are 6 base indices. In order to get things sequenced correctly, we simply use a sample sheet with a single dummy sample listed that has a fake 12 base index for index 1, and a fake 8 base index for index 2. What will happen is that the MiSeq will just sequence 12 bases for all reads for index 1, and 8 for index 2, regardless of their actual length or if index 2 is even present. After the run, everything will get thrown into the single Undetermined file and we demultiplex from there.

Now, there are a couple of options for demultiplexing, but I'll give the two we primarily use.

Our primary method is to use Qiime, which we use for our 16S analyses, to pull out the 16S and TruSeq libraries, as well as our Nextera libraries if we don't need the second index for discrimination. To do it this way, we figured out the 12 base versions of the 6 base TruSeq indices and 8 base Nextera indices and pass those 12 base versions to Qiime. This isn't ideal since Qiime won't do error correction for the TruSeq or Nextera indices, but it works fairly well and saves me from having to code my own solution. One thing to note here is that for our Nextera libraries, we have to make sure that we manually trim the adapters during analysis, because they're not trimmed off like they normally would be if using MiSeq Reporter.

The second method we use is to make a copy of the run folder, and then process it through MiSeq Reporter after modifying some of the output files. This is how we can demultiplex dual indexed Nextera samples. This is essentially the same process for how you would process a run that failed part-way through so that you could get some data out of it, but you're modifying the number of index cycles as opposed to a read. I generally don't like doing this method because it takes more space as we leave the original folder untouched and have to make a copy for reprocessing, which takes up a decent amount of space.

Either way, we've never seen issues of dataset contamination that can be traced back to running all of the samples together (ok, not entirely true but not for reasons related to the mixing of different index types) and/or our demultiplexing process.

[vstamborski]

We have run dual 8bp indexes and single 6bp indexes together, using the same trick as Calluna mentions for single vs dual. You simply always to enter the 8bp that would be sequenced as the index, even if it isn't really part of the index. For example, for the 6 bp TruSeq DNA index 1, you would enter ATC ACG AT and TCT TTCCC; for index 2 you would enter CGA TGT AT and TCT TTC CC, etc.