Run assembled genomes or metagenomes

Pass one .fna, .fa, or .fasta assembly, or place several assemblies in one directory. Each file stem becomes its sample identifier.

Run one assembly

pixi run ssuextract -q data/sample.fna --threads_per_job 4

Without --outdir, this command writes to results/sample.

Run a directory

pixi run ssuextract --query data/my_dataset --outdir results/my_dataset --threads_per_job 4

SSUextract validates query, model, database, and output paths before scheduling the workflow. Basenames may contain letters, numbers, ., _, and -.

Report more reference hits

Set the number of ranked BLAST subjects written for each extracted query:

pixi run ssuextract --query data/my_dataset --outdir results/my_dataset --top_hits 10

Inspect the rank, selection reason, public reference identifier, taxonomy, centroid taxonomy, identity, alignment length, and bit score:

cut -f1,4-5,7,9-10,17,19-20,28 results/my_dataset/blast_top_hits.tsv

The output contains the first 10 subjects, all equal-best assignment subjects, and the highest-ranked IMG, PR2, and SILVA subjects among the fetched candidates. The supplemental source rows do not change the assignment in cmsearch_summary.tsv.

Classify from tree neighbors

Enable tree classification for each extracted 16S rRNA gene or 18S rRNA gene:

pixi run ssuextract --query data/my_dataset --outdir results/my_dataset-tree --tree_classification --threads_per_job 4

Tree mode searches both marker indexes. The 100 best unique subjects determine whether the 16S rRNA gene or 18S rRNA gene covariance model is used for alignment. SSUextract aligns the query and references with cmalign, removes covariance-model insert columns and match columns with more than 90% gaps, then runs IQ-TREE 3 with GTR+F+R4, fast tree search, and 1,000 SH-aLRT replicates.

Inspect the selected tree taxonomy and the nearest reference branches:

cut -f1-3,29-53 results/my_dataset-tree/cmsearch_summary.tsv
cut -f1-11,13-20 results/my_dataset-tree/tree_nearest_neighbors.tsv

Per-query alignments, trees, model logs, reference FASTA files, and alignment QC are under results/my_dataset-tree/phylogeny/. The BLAST result remains in the blast_taxonomy fields and blast_top_hits.tsv.

If the selected marker yields fewer than three references, SSUextract keeps the BLAST taxonomy for that query and records the skipped tree attempt. Other queries continue.

Bound local resource use

Set Nextflow task ceilings independently from the per-search thread count:

pixi run ssuextract --query data/my_dataset --outdir results/my_dataset --threads_per_job 4 --max_cpus 8 --max_memory 64.GB

--threads_per_job controls each cmsearch, BLAST, and cmalign task. IQ-TREE uses one thread so near-tied branch distances are reproducible. --max_cpus limits the resources that Nextflow assigns to one task; it does not set the number of concurrent tasks.

Resume an interrupted run

Run the same command with Nextflow's resume flag:

pixi run ssuextract -resume --query data/my_dataset --outdir results/my_dataset

Keep the original work/ directory and command parameters. Nextflow reuses matching completed tasks and reruns missing work.