Query the KEGG KO HMMs in a single hmmsearch over a concatenated library

[edkerk]

Summary

getKEGGModelForOrganism previously ran one hmmsearch per KO (thousands of
invocations) against a directory of individual KO HMM files unpacked from
prok90_kegg116.zip / euk90_kegg116.zip. This PR replaces that loop with a
single hmmsearch of the whole query proteome against the concatenated
kegg116_<domain> HMM library, shipped as one gzip-compressed flatfile and
gunzipped on demand.

The profile library is the query and the proteome is the target sequence
database — the same fast search direction as before — so the reported per-hit
full-sequence E-values are identical to the historical per-KO search (same
profiles, same target db, same effective db size). The cut-off, the
minScoreRatioKO / minScoreRatioG filters and the downstream model assembly
are therefore unchanged; the thousands of hmmsearch calls simply collapse
into one.

Highlights:

• No hmmpress / hmmscan and no new bundled binary — hmmsearch already ships with RAVEN.
• Per-organism phylogenetic-distance subsampling (getPhylDist) is skipped for the prebuilt prok90/euk90 libraries, whose sequence set is already fixed — so the search path no longer needs a local keggdb/taxonomy.
• A legacy directory of per-KO HMMs is still honoured when already extracted (<dataDir>/hmms/K00844.hmm), so existing setups keep working.

What changed

reconstruction/kegg/getKEGGModelForOrganism.m:

• Acquisition — for the RAVEN-provided sets, prefer <dataDir>.hmm (the
  concatenated library); gunzip <dataDir>.hmm.gz if present, otherwise
  download kegg116_<domain>.hmm.gz. Falls back to the legacy .zip/per-KO
  directory when that is what the user has.
• Search + parse — when the concatenated library is used, run one
  hmmsearch --tblout and parse the table directly into koGeneMat
  (target col 1 = gene, query col 3 = KO, col 5 = full-sequence E-value),
  reusing the existing scoring/assembly downstream.

Validation

• MATLAB checkcode clean (no syntax errors; no new messages introduced).
• Search equivalence — on a mini library (8 KO profiles split out of
  prok90_kegg116) searched against a 300-sequence proteome, the per-KO
  hmmsearch loop and the single-library hmmsearch produce identical
  (gene, KO, E-value) hits (34 vs 34, zero differences).

Follow-up (release management)

The download URL points at
…/releases/download/v2.11.0/kegg116_<domain>.hmm.gz. The two library assets
(kegg116_prokaryotes.hmm.gz, kegg116_eukaryotes.hmm.gz) need to be attached
to the release before the download path is live. They are the same artefacts
published for raven-python.

[github-actions]

Function test results

0 tests   0 ✅  0s ⏱️
0 suites  0 💤
0 files    0 ❌

Results for commit a86e78a.