always report the rightmost match if multiple equivalent occurrences are present in target sequence? No. about vsearch HOT 5 CLOSED

Unfortunately, it is more complex, and the results depends on the rest of the sequences.

A counter-example: If the target/database sequence in the example had some extra non-matching sequence at the 3' end, the first match would be chosen. This is because these are global alignments and when a gap has to be opened at the 3' end anyway, it would rather extend that gap than opening an additional gap in 5' end, because the score would be better.

from vsearch.

tests added (frederic-mahe/vsearch-tests@b7beb67)

from vsearch.

@torognes it seems to me that the rule is to return the first perfect match found during backtracking? (so the rightmost one on the target sequence). If it is a permanent feature, I would like to mention it in the documentation. Is that ok for you?

Side question: I've been testing the same idea, but with matches on the minus strand of the target sequence. I've observed the same pattern (see frederic-mahe/vsearch-tests@70a6808), which suggests that when searching on the opposite strand, the query sequence is reverse-complemented and the target sequence is left unchanged. Is that correct?

from vsearch.

When aligning and backtracking, the code will always prefer to match the sequences than opening a gap, given that the scores are equal. It starts backtracking at the 3' ends of the sequences, and in the examples it will always start by matching the sequences at the 3' end.

from vsearch.

Thanks @torognes I've updated our test-suite (frederic-mahe/vsearch-tests@bae529a) with counter-examples. This is a complex behavior, and I don't see an easy way to describe it in the manpage, at least for now. That issue is complete and can be closed.

from vsearch.