DRAFT of AuxAlignment for parsing secondary/supplementary alignments

[clintval]

This is a draft PR to explore a common hierarchy for "alignments from SAM tags".

Motivated by:

• Add a class for parsing secondary alignments from XA/XB #206 (comment)

The diff is ugly but there are only 3 classes. I suggest reviewing like:

• AuxAlignment: abstract parent with some common concrete shared functions
• SecondaryAlignment: functionality specific to secondary alignments
• SupplementaryAlignment: functionality specific to supplementary alignments

I was able to make almost everything backwards compatible except:

• SupplementaryAlignment is now a dataclass vs attrs
• SupplementaryAlignment had a few init fields renamed

@tfenne what do you think?

I'm looking for a quick 👍 👎 so I can abandon the idea, or merge it into my actual feature branch for additional polish and a real round of review.

[codecov]

Codecov Report

Attention: Patch coverage is 88.88889% with 20 lines in your changes missing coverage. Please review.

Project coverage is 90.84%. Comparing base (b0b4227) to head (6d887a1).

Files with missing lines	Patch %	Lines
fgpyo/sam/__init__.py	88.88%	16 Missing and 4 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #208      +/-   ##
==========================================
- Coverage   91.06%   90.84%   -0.22%     
==========================================
  Files          18       18              
  Lines        2283     2437     +154     
  Branches      337      355      +18     
==========================================
+ Hits         2079     2214     +135     
- Misses        133      149      +16     
- Partials       71       74       +3     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[nh13]

I did not duplicate my comments from #206, so please incorporate those too as appropriate.

Instead of two sub-classes of AuxAlignment, what about storing an optional enum for if the alignment is secondary or supplementary (or None if neither or unknown)? Then for the from_tag(value: str) method, you could condition on if it's 4 values (XA), or 6 values with strand as the third item (SA), or 6 values with the cigar as the third item (XB). Alternatively, could you explain the motivation for the three classes beyond how they parse differently (which I think is resolved above)? Do we want to match on the class/type?

[nh13]

why not? I think a negative alignment score is aligned, especially in a global alignment.

[clintval]

I'd like to learn more about this concern! I haven't seen a score below zero for any output from bwa so I figured it would be a sign that data was corrupt. Couple questions:

1. Is there a way to run bwa in which the score can go below zero?
2. Knowing that this code will probably parse bwa tags for the forseeable future, do you think keeping the strict check is better for safety-sake or would hinder someone in the future with a popular or custom global aligner putting auxiliary alignments in SAM tags?

[clintval]

Comments from Nils in the other PR:

• Add a class for parsing secondary alignments from XA/XB #206 (review)
• Add a class for parsing secondary alignments from XA/XB #206 (comment)
• Add a class for parsing secondary alignments from XA/XB #206 (comment)
• Add a class for parsing secondary alignments from XA/XB #206 (comment)
• Add a class for parsing secondary alignments from XA/XB #206 (comment)
• Add a class for parsing secondary alignments from XA/XB #206 (comment)
• Add a class for parsing secondary alignments from XA/XB #206 (comment)
• Add a class for parsing secondary alignments from XA/XB #206 (comment)
• Add a class for parsing secondary alignments from XA/XB #206 (comment)

[clintval]

Closing this and will open a clean PR with a refactor.