[nucacidhunter]

I am just reading between the lines and assume that a pool of library was sequenced on both machines and a particular library yielded more reads on HiSeq than MiSeq. Some possible explanations:

1- different pooling ratio
2- differences in library size distribution
3- residual adapter in some libraries resulting in increased index hopping

I would do a demultiplexing of HiSeq reads with all index combinations to see if there has been unusual index hopping among some libraries (libraries with high index hopping will have less reads) and also pay attention to the number of unknown reads

[Innovelty]

On that note, what is an unusually high percentage of unknown reads?

[nucacidhunter]

Unknown reads source in indexed sequencing:

1- PhiX
2- indices that does not match those in sample sheet caused by base mis-match in excess of set numbers in bcl2 fastq. This is result of poor sequence quality due to over clustering or adapter/primer quality

Unknown reads are high when using desalted primers for amplicon libraries

[misterc]

IIRC, James Hadfield wrote up a very nice article on enseqlopedia showing how PhiX reads can get mis-associated with sample index sequences (when using Illumina's non-indexed PhiX) on patterned flow cells (like HiSeq 4K, X, and NovaSeq). If you do run PhiX with your samples on patterned flow cells at anything more than 1%, I'd suggest using indexed PhiX like the stuff SeqMatic sells. They'll even make you some PhiX with index sequences you specify (if desired).

[mama]

Misterc: could you give me the link of James Hadfield article?

[misterc]

http://enseqlopedia.com/2016/10/index-mis-assignment-to-illuminas-phix-control/

[HeinKey]

perhaps this question is asked before, but I don't understand the added value of dual indexing over single indexing anymore if both indices are unique in dual indexing. The highest possible complexity is then equal to highest complexity of single indexing.
What do I miss?

[nucacidhunter]

perhaps this question is asked before, but I don't understand the added value of dual indexing over single indexing anymore if both indices are unique in dual indexing. The highest possible complexity is then equal to highest complexity of single indexing.
What do I miss?

Your statement is correct but UDI addresses another issue (index hopping) in patterned flow cells used in HiSeq 3000/4000 HiSeqX and NovaSeq to maintain fidelity of reads assigned to multiplexed libraries. Links in post #1 should give you some information on the issue.

[Innovelty]

Sometimes, especially on the new Illumina flowcells, the indices get swapped around onto molecules they didn't originally belong to. With non-unique dual indices, where any combination of the index pool is potentially valid, that problem is invisible -- some of your reads *actually* belong to one library but have the indices for another library. With unique-dual-indexing, the problem still occurs, but it's not invisible -- any index-hopping or index-swapping will result in a read with an invalid combination of indices, and so it'll get kicked to the "Undetermined" set. You still "lost" that read, but it didn't get erroneously assigned to a different library.

Unique-dual-indexing *is* basically single-indexing, but you've got two barcodes, and the likelihood that both barcodes get swapped to something valid is practically nil.