MaskPrimers-align edge cases are not masking correctly

[Jason Vander Heiden]

When the alignment has a gap in the input sequence at the very end of the alignment, the masking is off by one. For example:

       ID> SRR765688.1679
SEQORIENT> RC
   PRIMER> LR11
 PRORIENT> F
  PRSTART> 33
    INSEQ> TCACCTGCGCTGTCTCTGGTGGCTCCATCAGCAGTAGTAACTGGTGGAGT-TGGGTCCGCAGCCC
    ALIGN> ---------------------------------GGTGCAGCTGGTGGAGTC
   OUTSEQ>                                  NNNNNNNNNNNNNNNNNTTGGGTCCGCAGCCC
   FIXSEQ>                                  NNNNNNNNNNNNNNNNNNTGGGTCCGCAGCCC

    ERROR> 0.2777777777777778

In this case, OUTSEQ has one extra T. Trivial attempts to fix the problem, solve the problem in the right-hand gap case, but introduce the problem in the left-hand gap case:

       ID> SRR765688.1837
SEQORIENT> RC
   PRIMER> LR3
 PRORIENT> F
  PRSTART> 0
    INSEQ> -GCAATCTGGGTCTGAGTTGAAGACGGCCTGGGGCCTCAGTGAAGATTTCCTGCAAGAC
    ALIGN> TGCAATCTGGGTCTGAGTTG-------------------------------
   OUTSEQ> NNNNNNNNNNNNNNNNNNNNAAGACGGCCTGGGGCCTCAGTGAAGATTTCCTGCAAGAC
   FIXSEQ> NNNNNNNNNNNNNNNNNNNNNAGACGGCCTGGGGCCTCAGTGAAGATTTCCTGCAAGAC
    ERROR> 0.050000000000000044

Will require more detailed parsing of the local alignments to fix both edge cases.

[ssnn]

Any chance that you already have some toy data to reproduce this and fix this bug?

[Jason Vander Heiden]

Not aside from the example I posted in the issue.

[Jason Vander Heiden]

I think the release this issue has sat for so long is that it’s probably a better use of time to trade out the function used for the local alignment, due to performance problems, instead of trying to hack a solution around the existing local alignment function.

[Jason Vander Heiden]

Look for striped smith waterman algorithm. The problem is finding a C implementation of it with a working python wrapper that is trivial to install (via pip). See the following for a starting point:

1. https://github.com/mengyao/complete-striped-smith-waterman-library
2. http://scikit-bio.org/docs/0.5.6/generated/skbio.alignment.StripedSmithWaterman.html#skbio.alignment.StripedSmithWaterman

‌

[Edel Aron]

I’ll definitely look into those, thanks Jason!

[Jason Vander Heiden]

Cool. There are a few more implementation out there. The developers of the first one are less than responsive:

https://github.com/mengyao/Complete-Striped-Smith-Waterman-Library/issues/55

But there have been multiple versions since I last tried to install it:

https://pypi.org/project/ssw-py/