Not all sub reads from multiple splits are written to FASTQ output file about duplex-tools HOT 4 CLOSED

Hi @callumparr,

The only reason a read would not be written out is if there's an overlapping match:

https://github.com/nanoporetech/duplex-tools/blob/master/duplex_tools/split_on_adapter.py#L173

We've decided to err on the side that doesn't cause any sequence duplication in the output in this case, but it's likely that we can solve it more elegantly somehow.

from duplex-tools.

Does overlapping match relate to the adapter sequences like below?

from duplex-tools.

You can see which spans are candidates to write out by looking at hits just before this line:

https://github.com/nanoporetech/duplex-tools/blob/master/duplex_tools/split_on_adapter.py#L164

The idea is to write out the complement of the adapter hits, so the following loops through the intervals against the adapter, and the hits will be populated with the complement of this (the end of each hit + the beginning of the next hit).

Note that 0,0 and len(seq),len(seq) are added to allow for the first and last "complement" to the adapter hits to be written out.

for left_hit, right_hit in pairwise(
        [(0, 0), *result['locations'], (len(seq), len(seq))]):
    hits.append([left_hit[1], right_hit[0]])

I'd imagine seeing what that variable is would bring some clarity into why there's one read that's not being written out

from duplex-tools.

Thank you. For now, it seems by eye that there is some overlap.

from duplex-tools.