GitHub
10/14/2026, 7:06 PMtestdata/mag_dereplicated/, a trimmed copy of an nf-core/mag output directory from a run with --dereplicate.
It has four bins in three clusters, with CheckM2, GTDB-Tk and Prokka results.
nf-core/magmap uses it to test --mag_outdir (nf-core/magmap#280).
The README section for the dataset says how it was made.
The three Prokka GFFs are about 1 MB each; the whole directory is 3 MB.
š¤ Generated with Claude Code
https://claude.ai/code/session_01UvGYU6kuXJeVv9ffdSzdzH
nf-core/test-datasetsGitHub
10/15/2026, 8:03 AM--mag_outdir. A few small comments below, none of them blocking.
What I checked ā
⢠All three dereplicated_bins/*.fa.gz files are valid gzip, and their contig IDs match the ##sequence-region entries of the corresponding Prokka GFFs exactly.
⢠dereplicated_bins.tsv gives three clusters as described in the README. MEGAHIT-MaxBin2-minigut.002 is only in the CheckM2 summary, which is a nice case to test.
⢠The GTDB-Tk summary has a species-level call, a partial lineage (g__;s__) and an Unclassified Bacteria row.
⢠The bin names change form between files: no extension in CheckM2, .fa in GTDB-Tk, .fa.gz in the dereplication files. Since that's what real mag output looks like, it's a good test of name handling in magmap.
Suggestions
1. GFF size (suggestion): about 85% of each GFF is the embedded ##FASTA section. If the files are cut at ##FASTA, the two large GFFs drop from ~1.1 MB to ~155 KB and ~227 KB, and the whole dataset shrinks from ~3 MB to ~1 MB. The sequences are already in dereplicated_bins/*.fa.gz, and as far as I can see magmap only passes a provided GFF to PROKKAGFF2TSV, so the embedded FASTA shouldn't be needed. Worth checking against the --mag_outdir implementation, though. Keeping the files unmodified for realism is also fine; in that case maybe just say so in the README.
2. README (minor): it says the three Prokka GFFs are about 1 MB each, but SPAdes-MaxBin2-test_minigut_sample2.001.gff is ~240 KB. It's two of the three.
3. Reproducibility (minor): since the data comes from 5.6.0dev plus an unreleased --bins_input branch, could you add the mag commit SHA/branch and the exact command or params file to the README? Then it can be regenerated if the layout changes once --bins_input is merged (e.g. the Annotation/Prokka/external/ path).
4. GTDB-Tk mock data (nit): SPAdes-MaxBin2-test_minigut.002 is classified as E. coli, but its closest placement is SelenomonasC. I assume this is an artefact of the mock-up database, which is fine for parsing tests. A one-line note in the README might save someone some confusion later.
Happy to approve once you've had a look. Thanks again! š
nf-core/test-datasetsGitHub
10/15/2026, 9:44 AM##FASTA (no stray lines, all 9-column records, contig IDs still match the bins). Dataset is down to ~1.1 MB.
⢠README size note updated, and the mag commit (f5218a59 on bins-input), params and the rerun note are a great addition for reproducibility.
⢠The mock-up GTDB-Tk note is clear.
⢠Bins, TSVs and summaries are byte-identical to the previous version.
LGTM, thanks for the quick turnaround! š
nf-core/test-datasetsGitHub
10/15/2026, 9:45 AM