add_sh_to_taxonomy.py (--addsh) miscounts rank columns when Species_exact is present

[erikrikarddaniel]

Description of the bug

bin/add_sh_to_taxonomy.py (used by --addsh to add UNITE species-hypothesis info onto DADA2's taxonomy table) determines which columns are taxonomic ranks with:

rank_cols = [
    c for c in taxtable.columns
    if c not in ("ASV_ID", "sequence", "confidence") and not c.endswith("_confidence")
]

This doesn't exclude Species_exact, written by DADA2_ADDSPECIES whenever assignTaxonomy's own taxlevels already reach species (the default, unless --skip_dada_addspecies is set). When --addsh and addSpecies are both active, Species_exact gets counted as a rank column, which throws off num_ranks:

• the SH column is inserted one position later than intended (after Species_exact instead of after Species)
• the per-rank values copied from the UNITE SH lookup (tax[1 : num_ranks + 1]) are misaligned by one column

Same root cause as the bug just fixed in bin/parse_dada2_taxonomy.r in #1062 (a hand-maintained non-rank-column exclusion list missing a column) -- found while reviewing that PR, but this is a separate script on a separate code path (--addsh), unrelated to #1062's own scope (DADA2 multi-database consolidation), so filing separately rather than folding the fix in there.

Possibly compounding this: existing --addsh test coverage (conf/test_pacbio_its.config) runs with --skip_dada_addspecies, specifically because addSpecies fails on those ITSx-extracted sequences for unrelated reasons -- so --addsh combined with addSpecies output doesn't currently have any test coverage at all.

Steps to reproduce

Run with --addsh and default addSpecies (i.e. without --skip_dada_addspecies) against a reference database whose taxlevels reach species, so assignTaxonomy already populates Species and DADA2_ADDSPECIES renames/duplicates it into Species_exact. Compare the SH-augmented taxonomy table's column order and values against the un-augmented table -- the SH column and the UNITE-lookup values should be shifted one column later than intended.

Suggested fix

Exclude Species_exact from rank_cols, e.g. by matching on suffix (_exact, mirroring the fix applied to bin/parse_dada2_taxonomy.r in #1062) rather than adding one more literal column name.