AlphaFold3 data pipeline never finds the Rfam database (Rfam-*.fasta glob does not match the shipped filename)

[SarahBeecroft]

Description of the bug

Hey,

I've found this bug and had claude write it up. see below:

RUN_ALPHAFOLD3_DATAPIPELINE always fails with ERROR: No Rfam database found, even when the Rfam database is present and was downloaded by the pipeline itself.

The module checks for the Rfam file with this glob:

https://github.com/nf-core/proteinfold/blob/dev/modules/local/run_alphafold3_datapipeline/main.nf#L69

rfam=$(ls -v ./rfam/Rfam-*.fasta 2>/dev/null | tail -n 1 || echo "")
if [[ -z "$rfam" ]]; then
    echo "ERROR: No Rfam database found"
    exit 1
fi

But the file the pipeline downloads is named rfam_14_9_clust_seq_id_90_cov_80_rep_seq.fasta. From conf/dbs.config:

alphafold3_rfam_link = 'https://storage.googleapis.com/alphafold-databases/v3.0/rfam_14_9_clust_seq_id_90_cov_80_rep_seq.fasta.zst'

Lowercase r, underscores, no hyphen. Rfam-.fasta cannot match it on a case-sensitive filesystem. Nothing between ARIA2_RFAM in PREPARE_ALPHAFOLD3_DBS and the module renames the file, and the publish pattern in conf/modules_alphafold3.config is rfam.fasta, which confirms the lowercase name is what lands in ${params.alphafold3_db}/rfam/.

This is the same name DeepMind's own AlphaFold3 distribution uses (see src/alphafold3/scripts/copy_to_ssd.sh in google-deepmind/alphafold3), so users supplying a pre-existing AF3 database directory via --alphafold3_db hit it too.

This looks like a regression introduced by commit 357402c8 ("Fix path to check for rfam dbs in af3 datapipeline module", 28 Jul 2026), which changed the glob in the wrong direction:

diff
-    rfam=$(ls -v ./rfam/rfam*.fasta 2>/dev/null | tail -n 1 || echo "")
+    rfam=$(ls -v ./rfam/Rfam-*.fasta 2>/dev/null | tail -n 1 || echo "")

Rfam-*.fasta matches the HelixFold3 naming (Rfam-14.9_rep_seq.fasta, used in modules/local/run_helixfold3/main.nf), not the AlphaFold3 one.

The adjacent checks in the same module are fine: nt_rna*.fasta and rnacentral*.fasta both match their downloaded filenames. Rfam is the only mismatched one.

Suggested fix

Revert the glob, or accept both spellings:

diff
-    rfam=$(ls -v ./rfam/Rfam-*.fasta 2>/dev/null | tail -n 1 || echo "")
+    rfam=$(ls -v ./rfam/[Rr]fam[-_]*.fasta 2>/dev/null | tail -n 1 || echo "")

Command used and terminal output

Relevant files

No response

System information

No response