Methods

Methods

What each step runs, with the exact settings that are not defaults.

Long reads

1. Trim (opt-in, --trim-long): Porechop removes leftover adapters and splits chimeric reads at middle adapters (--check_reads 1000). Off by default: recent basecallers (Dorado) trim adapters themselves and Porechop is unmaintained and slow — reach for it on older reads, or when the report's read QC shows adapter artefacts.
2. Filter (on by default, --no-filter-long skips): Filtlong keeps the best 95% of the bases (--keep_percent 95, --min_length 500); with --size, at most the best 100x of the genome (--target_bases).
3. Assemble: Flye (--nano-hq, 3 polishing iterations; expects recent super-accuracy basecalls) or Shasta (--config Nanopore-May2022, minimum read length 3,000 unless --min-size). The assembly graph is kept and drawn with Bandage; per-sample stats land in flye_stats.tsv / shasta_stats.tsv.
4. Polish, long reads: Medaka (medaka_consensus) with the model from --model. The default is ONT's bacterial methylation model, the best performer in our benchmarks; for non-bacterial samples or other flowcells pick the model matching your basecaller — a wrong model adds errors. --long-polisher none skips this: do that for recent super-accuracy basecalls at high coverage (little left to fix), and always for non-Nanopore long reads (Medaka on PacBio doubled the mismatch rate in our benchmark).

Short reads (hybrid polishing)

5. Trim (on by default, --no-trim-short skips): fastp with a sliding-window quality cut (--cut_right, mean quality 10) and a 64 bp length floor; adapters auto-detected for pairs. With --ion-torrent, adapter trimming is disabled (the Torrent Suite removes adapters before export, and fastp's detection is tuned to Illumina chemistry); the quality window stays, since Ion quality also decays toward the 3' end.
6. Polish, short reads: Polypolish (BWA -a all-alignments mapping of each mate, as its author specifies), then pyPOLCA with --careful. One Polypolish round only: a second round changed nothing in our tests. This pair was chosen because neither introduces errors in low-coverage regions — NextPolish, ntEdit and plain POLCA did, and were dropped. The final contigs are renamed contig_1, contig_2, … Polish only with Illumina reads: Ion Torrent reads share Nanopore's homopolymer error mode — in our benchmark they fixed ~92% of the indels but 96–97% of what they left behind were homopolymer indels, where depth-matched Illumina left none. With --ion-torrent HashBrown warns accordingly.

Short reads only

• Assemble: SKESA (conservative, recommended for Illumina isolates) or SPAdes (scaffolds; --ion-torrent adds SPAdes' Ion Torrent mode; single-end supported by both). SKESA graphs are rebuilt with gfa_connector when it is installed.

Reorient (every run, optional)

• dnaapler (dnaapler all), when installed, rotates every final assembly to the conventional start: the chromosome at dnaA, plasmids at their rep gene. The reorientation happens in place and the pre-reorientation sequence is kept as <sample>_unoriented.fasta. The dot plots always compare pre-reorientation sequences, so a rotation never masquerades as a structural change.

QC (every run)

• QUAST over the final assemblies (--min-contig 1000).
• Mapping QC per sample, read type and read-processing stage (raw, trimmed, filtered — every stage the run produced), on a minimap2 mapping (map-ont or sr preset, unmapped reads dropped, duplicates removed with samtools markdup): samtools flagstat for the mapping rates and mosdepth for the depth. Comparing stages shows what trimming/filtering bought (mapping rate up, depth cost).
• Coverage: mosdepth's mean depth over the final assembly, one table per read type.
• Dot plots: nucmer + mummerplot comparing the assembly before and after each polishing stage (raw vs Medaka, Medaka vs short-read polished) — a quick visual check that polishing did not rearrange anything.
• Assembly graphs: Bandage PNGs of the Flye/Shasta/SPAdes/gfa_connector graphs.

Report (every run)

Every run ends with a per-sample HTML report (plus a PDF twin and a run-level index — see Outputs). Its core needs nothing beyond the HashBrown environment:

• Read funnel per read type: seqkit statistics of every stage (raw → trimmed → filtered), with per-stage length/quality distributions and, when pycoqc-ng is pip-installed, a full pycoQC-ng page per stage (its converters rebuild the sequencing summary from fastq, for Nanopore and Illumina alike).
• Assembly funnel: contigs, size, N50, %GC — and, with Merqury installed, the consensus QV of every polishing stage from the short-read 21-mers (reference-free; the in-run version of our validation measurements).
• Species confirmation: mash against a local RefSeq sketch (hashbrown db mash), the top hit named and fetched with NCBI datasets, ANI with fastANI or skani (same species: ≥95%).
• Completeness & sanity: CheckM2 completeness/contamination, BUSCO (prokaryote auto-lineage), barrnap rRNA counts, tRNAscan-SE tRNA counts.
• Typing & genes: MLST, MOB-suite plasmid reconstruction and typing, and one merged gene table from AMRFinderPlus (resistance), abricate+VFDB (virulence) and abricate+PlasmidFinder (replicons).
• Annotation: Bakta by default when installed (--annotate), NCBI PGAP via Docker on request.
• Contamination: ConFindr on the short reads, a GenomeScope2 read-based genome profile, and a blob plot — every contig by %GC and coverage, dot area ~ length, colored by its closest RefSeq genome (per-contig mash dist -i).

QC criteria and the MIMAG verdict

The report's "QC criteria" table judges every metric against an explicit threshold and shows the verdict as a pass / warn / fail chip; the thresholds anchor on the MIMAG high-quality draft standard (Bowers et al. 2017) where one exists, plus pipeline-sensible read-coverage minimums:

Metric	Pass	Warn	Fail	Source
CheckM2 completeness	> 90%	50–90%	< 50%	MIMAG
CheckM2 contamination	< 5%	5–10%	≥ 10%	MIMAG
rRNA genes	≥ 1 each of 5S, 16S, 23S	—	any missing	MIMAG
tRNA isotypes	≥ 18	< 18	—	MIMAG
Long-read coverage	≥ 30x	< 30x	—	pipeline
Short-read coverage (when present)	≥ 30x	< 30x	—	pipeline
ANI to closest genome	≥ 95%	—	< 95%	species boundary (Jain et al. 2018)
ConFindr	Clean	—	Contaminated	ConFindr

The headline verdict is the MIMAG draft-quality tier: high-quality needs completeness > 90%, contamination < 5%, all three rRNAs and ≥ 18 tRNA isotypes; medium-quality needs completeness ≥ 50% and contamination < 10%; anything else is low-quality. A metric whose module did not run shows as "not assessed" — it never counts as a pass, and without CheckM2 no tier is assigned.

Each of these modules is optional: a missing tool or database turns into a "skipped (install hint)" note in the report, never an error. The Usage page lists the install one-liners and hashbrown db downloads. The version of every program used is written to log.txt in the output folder.

References

• Flye — Kolmogorov M, et al. Assembly of long, error-prone reads using repeat graphs. Nat Biotechnol 37, 540–546 (2019).
• Shasta — Shafin K, et al. Nanopore sequencing and the Shasta toolkit enable efficient de novo assembly of eleven human genomes. Nat Biotechnol 38, 1044–1053 (2020).
• Medaka — Oxford Nanopore Technologies, https://github.com/nanoporetech/medaka
• Porechop — Wick RR, https://github.com/rrwick/Porechop
• Filtlong — Wick RR, https://github.com/rrwick/Filtlong
• SKESA (and gfa_connector) — Souvorov A, et al. SKESA: strategic k-mer extension for scrupulous assemblies. Genome Biol 19, 153 (2018). https://github.com/ncbi/SKESA
• SPAdes — Prjibelski A, et al. Using SPAdes de novo assembler. Curr Protoc Bioinformatics 70, e102 (2020).
• fastp — Chen S, et al. fastp: an ultra-fast all-in-one FASTQ preprocessor. Bioinformatics 34, i884–i890 (2018).
• Polypolish — Wick RR, Holt KE. Polypolish: short-read polishing of long-read bacterial genome assemblies. PLoS Comput Biol 18, e1009802 (2022).
• pyPOLCA — Bouras G, et al. How low can you go? Short-read polishing of Oxford Nanopore bacterial genome assemblies. Microb Genom 10, 001254 (2024). POLCA: Zimin AV, Salzberg SL. PLoS Comput Biol 16, e1007981 (2020).
• BWA — Li H, Durbin R. Fast and accurate short read alignment with Burrows–Wheeler transform. Bioinformatics 25, 1754–1760 (2009).
• minimap2 — Li H. Minimap2: pairwise alignment for nucleotide sequences. Bioinformatics 34, 3094–3100 (2018).
• SAMtools — Danecek P, et al. Twelve years of SAMtools and BCFtools. GigaScience 10, giab008 (2021).
• QUAST — Gurevich A, et al. QUAST: quality assessment tool for genome assemblies. Bioinformatics 29, 1072–1075 (2013).
• mosdepth — Pedersen BS, Quinlan AR. Mosdepth: quick coverage calculation for genomes and exomes. Bioinformatics 34, 867–868 (2018).
• MUMmer4 — Marçais G, et al. MUMmer4: a fast and versatile genome alignment system. PLoS Comput Biol 14, e1005944 (2018).
• Bandage — Wick RR, et al. Bandage: interactive visualization of de novo genome assemblies. Bioinformatics 31, 3350–3352 (2015).
• seqkit — Shen W, et al. SeqKit2: a Swiss army knife for sequence and alignment processing. iMeta 3, e191 (2024).
• pycoQC-ng — Duceppe MO, https://github.com/duceppemo/pycoqc-ng (fork of pycoQC: Leger A, Leonardi T. J Open Source Softw 4, 1236 (2019)).
• dnaapler — Bouras G, et al. Dnaapler: a tool to reorient circular microbial genomes. J Open Source Softw 9, 5968 (2024).
• Merqury — Rhie A, et al. Merqury: reference-free quality, completeness, and phasing assessment for genome assemblies. Genome Biol 21, 245 (2020).
• Mash — Ondov BD, et al. Mash: fast genome and metagenome distance estimation using MinHash. Genome Biol 17, 132 (2016).
• FastANI — Jain C, et al. High throughput ANI analysis of 90K prokaryotic genomes reveals clear species boundaries. Nat Commun 9, 5114 (2018).
• CheckM2 — Chklovski A, et al. CheckM2: a rapid, scalable and accurate tool for assessing microbial genome quality using machine learning. Nat Methods 20, 1203–1212 (2023).
• BUSCO — Manni M, et al. BUSCO update: novel and streamlined workflows. Mol Biol Evol 38, 4647–4654 (2021).
• barrnap — Seemann T, https://github.com/tseemann/barrnap
• tRNAscan-SE — Chan PP, et al. tRNAscan-SE 2.0: improved detection and functional classification of transfer RNA genes. Nucleic Acids Res 49, 9077–9096 (2021).
• mlst — Seemann T, https://github.com/tseemann/mlst (uses PubMLST: Jolley KA, et al. Wellcome Open Res 3, 124 (2018)).
• MOB-suite — Robertson J, Nash JHE. MOB-suite: software tools for clustering, reconstruction and typing of plasmids from draft assemblies. Microb Genom 4, e000206 (2018).
• AMRFinderPlus — Feldgarden M, et al. AMRFinderPlus and the Reference Gene Catalog. Sci Rep 11, 12728 (2021).
• abricate — Seemann T, https://github.com/tseemann/abricate (VFDB: Liu B, et al. Nucleic Acids Res 50, D912–D917 (2022); PlasmidFinder: Carattoli A, et al. Antimicrob Agents Chemother 58, 3895–3903 (2014)).
• Bakta — Schwengers O, et al. Bakta: rapid and standardized annotation of bacterial genomes via alignment-free sequence identification. Microb Genom 7, 000685 (2021).
• PGAP — Tatusova T, et al. NCBI prokaryotic genome annotation pipeline. Nucleic Acids Res 44, 6614–6624 (2016).
• ConFindr — Low AJ, et al. ConFindr: rapid detection of intraspecies and cross-species contamination in bacterial whole-genome sequence data. PeerJ 7, e6995 (2019).
• GenomeScope2 — Ranallo-Benavidez TR, et al. GenomeScope 2.0 and Smudgeplot for reference-free profiling of polyploid genomes. Nat Commun 11, 1432 (2020).
• WeasyPrint — Kozea, https://weasyprint.org
• MIMAG — Bowers RM, et al. Minimum information about a single amplified genome (MISAG) and a metagenome-assembled genome (MIMAG) of bacteria and archaea. Nat Biotechnol 35, 725–731 (2017).