bgcflow completeness audit (2026-08-21)
Upstream: NBChub/bgcflow @ v1.1.2 · Port: oxo-flow-bgcflow (live-verified).
Seven runnable Snakefiles — the most entry-point-rich upstream of the batch (plus a broken Report mode):
| snakefile | what it runs |
|---|---|
| main Snakefile | prokka annotation backbone + gtdb meta + mibig table + ~20 toggle pipelines (eggnog, mash, fastani, automlst, roary, bigslice, checkm, gtdbtk, antismash, arts, deeptfactor, cblaster×2, bigscape, gecco, amrfinderplus, seqfu) |
| BGC | BGC-id-driven selection + bigslice/bigscape/clinker/interproscan/mmseqs2/getphylo/automlst |
| Database | dbt+DuckDB warehouse builder (requires main outputs; dbt repo clone; cargo asdb-taxa) |
| Metabase | metabase jar + duckdb plugin + interactive server (hardcoded admin creds!) |
| Alleleome | hidden entry (not in wrapper list): roary → alleleome core-gene fasta |
| lsabgc | lsaBGC autoanalyze on main intermediates (Kofam OFF by default) |
| ppanggolin | 2× 15-rule ppanggolin ladders (genome + roary-based) |
Gap tiers vs the port
P0: the other 6 snakefiles (BGC/Database/Alleleome/lsabgc/ ppanggolin — Metabase = P2 infra with a hardcoded-credentials footgun); the ~15 non-ported pipelines in the main toggle set (bigslice/query-bigslice/bigscape/clinker/interproscan/mmseqs2/ arts/deeptfactor×2/cblaster×2/gecco/amrfinderplus/roary/eggnog/ automlst/seqfu/mash/fastani); sample-source branches (ncbi/patric/ gbk-convert).
P1: KEGG Kofam data (license; OFF by default upstream — keep off + document); InterProScan EULA (implicit in direct download).
P2: antismash v6/v7 switch, gtdbtk release/ani_screen, use_gtdb_api offline mode, BGCFLOW_ANTISMASH_MODE env toggle, resources_path symlinks.
Data gates (all free, some huge): GTDB API/metadata (~60-90GB for gtdbtk), MIBiG JSON, antiSMASH DBs, eggNOG bacteria.dmnd, CheckM data, BiG-SCAPE+Pfam-A, BiG-FAM (query-bigslice), ARTS refs, AMRFinderPlus DB, lsaBGC annotation DBs, NCBI/PATRIC downloads, GECCO models, DeepTFactor clone.
Upstream dead code (do not port): Report mode broken (dir not snakefile); report.smk included by nothing; mlst/refseq_masher rules unreachable via rule all; antismash-db-duckdb only in Database mode.
Verdict
Largest multi-entry surface — 7 snakefiles × ~20 pipelines. The ported default main path (prokka + antismash + gtdb) leaves the BGC-selection mode and the warehouse/reporting stack as the biggest P0 blocks.
Re-verification (2026-08-23, 9-mini queue 1/9)
Engine: latest main (post-v0.14.1) · Box: bioinfo-wsx · Mode: real CLI
run, not dry-run · -j 4 --keep-going.
14 succeeded, 13 skipped, 0 failed, exit 0, one round, zero repo fixes. Coverage chain: copy_custom_fasta → gtdb_prep (bac120 metadata pre-staged via Mac relay) → prokka ×2 → format_gbk ×2 → get_mibig_table → antismash 7 full DB set (pfam 293MB + clusterblast 751MB + 8 more components, pre-staged and sha256-accepted) → antismash_genomes S1+S2 → overview/copy/summary → csv_to_parquet. All 13 skips are checkpoint reuse.