dānaSeq nanopore_assembly and illumina_assembly: long- and short-read metagenome assembly

Motivation: Metagenome-assembled genomes are binned from contigs and per-sample coverage profiles, whose production takes most of the compute time of an environmental metagenomics project. dānaSeq is a suite of six Nextflow pipelines from the Cryomics Lab; this paper describes its two assembly pipelines. Results: nanopore_assembly takes Oxford Nanopore reads from a MinKNOW barcode tree or a flat directory, deduplicates and optionally size-selects them with fastq_filter, an order-independent read selector, removes human reads, co-assembles with Flye, metaMDBG or myloasm, optionally polishes, and maps every sample back with minimap2 for a CoverM depth table. illumina_assembly takes paired-end Illumina reads, runs BBTools quality control and error correction, assembles each sample or a co-assembly with up to four assemblers, deduplicates the pooled contigs and maps reads back with BBMap. Both emit an assembly FASTA, a MetaBAT-format depth table and indexed BAM files, the inputs of mag_analysis. Both run from conda environments or Docker/Apptainer images, on a workstation or inside a Slurm allocation, with node-local scratch and stage caching for resubmission. On the Grex cluster, fastq_filter (earlier revision) selected 50% of the bases of 0.1–10 Gbp read sets 3.63–3.66 times faster than Filtlong v0.3.1, with 6.0–6.5 MB peak memory. With earlier commits, a Flye co-assembly of 14.1 GiB of gzipped reads took 1 h 23 min on 64 cores, and full-size co-assemblies took 28–41 h on 160 cores. Availability and implementation: dānaSeq is released under the MIT license at https://github.com/rec3141/danaSeq (this description corresponds to release v1.3.0, https://github.com/rec3141/danaSeq/releases/tag/v1.3.0); documentation at https://rec3141.github.io/danaSeq/.