<#2255 feat: add per-sample assemblies to the meta...
# test-data
g
#2255 feat: add per-sample assemblies to the metagenome binning dataset Pull request opened by dialvarezs
SemiBin2 multi_easy_bin
and
Vamb
, and co-assembly binners in general, need one assembly per sample combined into a single FASTA with the sample as a header prefix. The binning dataset added in #2237 shares one assembly across its three samples, so it cannot drive that mode, and nf-core/modules still tests
semibin/multieasybin
against data under
delete_me/
. This adds
binning/multisample/
, covering the same community in that form. It is derived entirely from the files already in
binning/
, so there is no second simulation to keep in step: • the reads come back out of the published BAMs with
samtools fastq
, split by genome on the reference name • the contigs are the published 5 kb pieces stitched into blocks of 8, with each sample taking a seeded 60% of them, which keeps the contig count near what a real assembly would give • each sample takes one genome's reads from one published sample, giving five abundance ratios out of the three that exist Contents are
contigs.fasta.gz
,
s1
to
s5
.sorted.bam
with indexes,
s1
to
s5
.aemb.tsv
and
abundances.tsv
. It carries five samples rather than three because SemiBin2 only accepts
--abundance
in place of BAMs from five samples up. Both tools' abundance forms are included, since they differ: SemiBin reads one file per sample over the split contigs
SemiBin2 split_contigs
produces, Vamb one wide table over the contigs themselves. The separator is
C
, which is Vamb's default and what SemiBin takes as
-s C
. Using SemiBin's own default of
:
works for both tools but leaves a colon in the names of the bins Vamb writes. Verified on all four paths. SemiBin2 2.4.1 returns 2, 2, 3, 2 and 2 bins from the BAMs and 2 per sample from the abundance files, about 70 seconds a run. Vamb 5.0.4 returns 53 clusters in 163 split bins from either input, at the low epoch count nf-core/modules uses for testing. The README section documents the full recipe and the seeds. Draft until nf-core/modules#12863 is pointed at it. 🤖 Generated with Claude Code nf-core/test-datasets
Review on #2255 feat: add per-sample assemblies to the metagenome binning dataset Review approved by erikrikarddaniel on dialvarezs's pull request 👍 nf-core/test-datasets