Indexes
Why Hamming distance 1 between Illumina indexes is a failed pool
Demultiplexing needs leftover substitutions after an error. Distance 1 cannot provide that, and combinatorial dual indexes add an index-hopping failure even when the pair looks unique.
What demultiplexing is actually comparing
Illumina assigns a read to a sample by comparing the sequenced index to the barcodes in the sample sheet. Hamming distance is the number of substitutions between two sequences of the same length. If two samples in the lane are ATCACGTT and ACCACGTT, the distance is 1. One sequencing error on the first barcode produces the second.
The usual rule is a minimum distance of 3 in a pool. Then a single substitution still leaves the read closer to the true index than to any other. Distance 2 is already fragile. Distance 1 is not a pool you can rescue with quality filters.
Unique dual indexes are a different failure
Dual indexing uses i7 and i5. The demultiplexer keys on the pair, so two samples can share an i7 if their i5 sequences differ. That is combinatorial dual indexing. On patterned flow cells (NovaSeq, HiSeq 4000, and related instruments) index hopping can attach the wrong i5 to a fragment, and the hopped pair can match another sample in the sheet.
Unique dual indexes avoid that: every i7 is unique in the pool, and every i5 is unique. Hamming distance on each oligo still applies. A pool can pass pair uniqueness and still fail unique-dual or distance checks. Those are separate findings.
What we check
The checker reads index tables in your browser. It reports exact duplicates, pairs below the Hamming threshold you set, reverse complements (a common i5 orientation mix-up), and combinatorial i7 or i5 reuse. It does not look up catalog names, rewrite a sample sheet, or simulate hopping rates. The export records the version and the settings that produced the report.