fix: six bugs that break FORMAT_KEGG_DB and KEGG annotation pipeline

[ruben1294]

Summary

This PR fixes six bugs found while running the KEGG database formatting
and annotation pipeline (FORMAT_KEGG_DB process). Together they cause
the pipeline to fail when format_kegg = true or produce
silently wrong results. All fixes were validated end-to-end on an HPC
cluster running Nextflow with Docker.

---

Bug 1 — format_kegg_database.py: top-level scikit-bio import crashes containers without it

File: bin/format_kegg_database.py

from skbio import ... was imported at module level. Any container that does not have scikit-bio installed raises ImportError on startup, before any real work is done. scikit-bio is only needed when a gene_ko_link file is actually processed, so the imports are now moved into the two functions that require them: process_kegg and generate_modified_kegg_fasta.

---

Bug 2 — format_kegg_database.py: MMseqs2 DB written as kegg.<date>.mmsdb, expected as kegg.mmsdb

File: bin/format_kegg_database.py

process_kegg() previously wrote the MMseqs2 database as:

kegg_mmseqs_db = path.join(output_dir, "kegg.%s.mmsdb" % download_date)
# e.g. kegg/kegg.2018-12-30.mmsdb

However, modules/local/annotate/mmseqs_search.nf symlinks the database directory and then references the database as ${db_name}.mmsdb (where db_name is the directory name "kegg"), so it expects exactly kegg/kegg.mmsdb. The date suffix caused a silent "file not found" error at annotation time.

This is fixed by removing the date suffix:

kegg_mmseqs_db = path.join(output_dir, "kegg.mmsdb")

---

Bug 3 — format_kegg_database.py: --skip_gene_ko_link argparse uses type=bool instead of action="store_true"

File: bin/format_kegg_database.py

type=bool does not work as a flag. argparse passes the literal string "False" to bool(), and bool("False") == True, so any invocation with --skip_gene_ko_link would receive the wrong value.

# Before (broken):
parser.add_argument("--skip_gene_ko_link", type=bool, default=False, ...)

# After (fixed):
parser.add_argument("--skip_gene_ko_link", action="store_true", ...)

---

Bug 4 — format_kegg_db.nf: container python_scikit-bio_scipy does not include mmseqs2

File: modules/local/database/format_kegg_db.nf

The FORMAT_KEGG_DB process called mmseqs createdb and mmseqs createindex
but ran inside python_scikit-bio_scipy, which does not ship mmseqs2.
This caused an immediate FileNotFoundError: mmseqs.

The container is now aligned with the rest of the pipeline and uses an image that includes mmseqs2:

// Before:
container "community.wave.seqera.io/library/python_scikit-bio_scipy:0f89a100e990daf2"

// After:
container "community.wave.seqera.io/library/python_pandas_hmmer_mmseqs2_pruned:d2c88b719ab1322c"

---

Bug 5 — format_kegg_db.nf: bash condition if [ ${skip_gene_ko_link} ] is always true

File: modules/local/database/format_kegg_db.nf

In bash, if [ "0" ] evaluates to true because any non-empty string is truthy. As a result, the process always took the --skip_gene_ko_link branch and silently ignored the gene_ko_link file even when it was provided.

This is fixed by comparing explicitly to the string "true" (see also Bug 6):

# Before:
if [ ${skip_gene_ko_link} ]; then

# After:
if [ "${skip_gene_ko_link}" = "true" ]; then

---

Bug 6 — dram.nf: skip_gene_ko_link passed as integer 1/0 instead of string "true"/"false"

File: workflows/dram.nf

The value passed to FORMAT_KEGG_DB was:

// Before:
skip_gene_ko_link = params.skip_gene_ko_link ? 1 : 0

This produces the strings "1" or "0" inside the bash script. As noted in Bug 5, if [ "0" ] is true in bash. The fix in format_kegg_db.nf requires the value to be exactly "true" or "false":

// After:
skip_gene_ko_link = params.skip_gene_ko_link ? "true" : "false"

---

Files changed

File	Bugs fixed
bin/format_kegg_database.py	#1 (lazy scikit-bio import), #2 (kegg.mmsdb name), #3 (argparse flag)
modules/local/database/format_kegg_db.nf	#4 (container with mmseqs2), #5 (bash condition)
workflows/dram.nf	#6 (boolean string value)