Nanopore Assembly Pipeline¶

Long-read metagenomic assembly pipeline for Oxford Nanopore data. Preprocesses reads, runs assembly with selectable assemblers, maps reads back to contigs, and produces depth tables for downstream MAG analysis.

Quick Start¶

cd nanopore_assembly
./install.sh && ./install.sh --check

# Local (conda)
./run-nanopore-assembly.sh --input /path/to/reads --outdir /path/to/output

# Choose assembler
./run-nanopore-assembly.sh --input /path/to/reads --outdir /path/to/output --assembler metamdbg

# Apptainer (HPC)
./run-nanopore-assembly.sh --apptainer --input /path/to/reads --outdir /path/to/output

# Docker
./run-nanopore-assembly.sh --docker --input /path/to/reads --outdir /path/to/output

Pipeline Stages¶

Step	Process	Tool	Description
1	`CONCAT_READS`	cat	Concatenate per-barcode FASTQs into one file per sample
2	`PREPARE_READS`	BBDuk / Filtlong	Deduplication, size-selection, quality filtering
3	`REMOVE_HUMAN`	minimap2	Remove human-mapped reads (optional)
4a	`FLYE_ASSEMBLE`	Flye --meta	Long-read metagenome assembly (default)
4b	`ASSEMBLY_METAMDBG`	metaMDBG	Minimizer-space de Bruijn graph assembly (alternative)
4c	`ASSEMBLY_MYLOASM`	myloasm	Hybrid overlap/DBG assembly (alternative)
5	`FLYE_POLISH`	Flye	Assembly polishing iterations (optional, default for Flye)
6	`MAP_READS`	minimap2 + samtools	Per-sample alignment to assembly (flag `-F 0x904`)
7	`CALCULATE_DEPTHS`	CoverM	Coverage depth table in MetaBAT2 format
8	`CALCULATE_TNF`	tetramer_freqs (C)	136-feature tetranucleotide frequency profiles

Parameters¶

Parameter	Default	Description
`--input`	(required)	Directory containing `.fastq.gz` or `fastq_pass/barcode/` structure
`--outdir`	`results`	Output directory
`--assembler`	`flye`	Assembler: `flye`, `metamdbg`, or `myloasm`
`--read_type`	`auto`	Nanopore read type: `auto`, `nano-hq`, `nano-raw`, `nano-corr`
`--min_overlap`	`1000`	Flye `--min-overlap` parameter
`--polish`	(auto)	Run Flye polisher; auto-enabled for Flye, disabled for others
`--dedupe`	`true`	BBDuk deduplication before assembly
`--dedupe_memory`	`24 GB`	Memory allocated to deduplication
`--filtlong_size`	(skip)	Filtlong target bases (e.g. `40000000000`); skip if not set
`--min_barcode_size`	`10485760`	Skip barcodes with fewer than 10 MB of reads
`--run_remove_human`	`true`	Remove human reads via minimap2
`--human_ref`	`databases/human_ref/GRCh38_noalt_as.fa.gz`	Path to human reference FASTA or `.mmi`
`--assembly_cpus`	`16`	CPUs for assembly processes
`--assembly_memory`	`60 GB`	Memory for assembly processes
`--store_dir`	(none)	Persistent cache directory (storeDir)

Outputs¶

results/
├── assembly/
│   ├── assembly.fasta          Co-assembly
│   ├── assembly_info.txt       Contig metadata (length, coverage, circularity)
│   ├── assembly_graph.gfa      Assembly graph
│   ├── tnf.tsv                 Tetranucleotide frequencies (136 features)
│   └── gc.tsv                  Per-contig GC content
├── mapping/
│   ├── <sample>.sorted.bam     Per-sample alignments
│   ├── <sample>.sorted.bam.bai BAM indices
│   └── depths.txt              CoverM depth matrix (MetaBAT2 format)
└── pipeline_info/
    ├── timeline.html           Execution timeline
    ├── report.html             Resource usage report
    ├── trace.txt               Per-process trace
    ├── dag.html                Pipeline DAG
    └── versions.yml            Tool versions and parameters

Feed these outputs into mag_analysis:

cd ../mag_analysis
./run-mag-analysis.sh \
    --assembly /path/to/results/assembly/assembly.fasta \
    --depths /path/to/results/mapping/depths.txt \
    --bam_dir /path/to/results/mapping/ \
    --outdir /path/to/analysis --db_dir /path/to/databases

Profiles¶

Profile	Use case
`standard`	Local execution (default)
`test`	Small test data, reduced resources (4 CPUs, 8 GB)

Resource Requirements¶

Component	CPUs	RAM	Notes
Flye assembly	16	60 GB	Scales with metagenome complexity
metaMDBG assembly	16	60 GB	Lower memory than Flye for large datasets
myloasm assembly	16	60 GB	Hybrid overlap/DBG approach
minimap2 mapping	8	16 GB	Per-sample; runs in parallel
CoverM depths	2	4 GB	Operates on all BAMs
Human decontamination	8	16 GB	minimap2 against GRCh38

Input¶

Nanopore barcode directory structure (MinKNOW output):

input_dir/
├── fastq_pass/
│   ├── barcode01/*.fastq.gz
│   ├── barcode02/*.fastq.gz
│   └── ...

Or a flat directory of FASTQ files:

input_dir/
├── sample1.fastq.gz
├── sample2.fastq.gz
└── ...

Barcodes with fewer than --min_barcode_size bytes (default 10 MB) are automatically skipped.

Design Notes¶

Multiple assemblers. Flye (default), metaMDBG, and myloasm each handle different read quality profiles. Select with --assembler.
CoverM depth calculation. Uses CoverM instead of MetaBAT2's jgi_summarize_bam_contig_depths to avoid integer overflow with supplementary alignments.
Human decontamination. Enabled by default via minimap2 against GRCh38. Disable with --run_remove_human false.
Tetranucleotide frequencies. 136-feature TNF profiles computed by a compiled C binary for downstream binning and visualization.
Persistent caching. Use --store_dir to skip completed processes across runs, even after work/ cleanup.
Auto-polishing. Polishing is automatically enabled for Flye assemblies and disabled for metaMDBG/myloasm. Override with --polish true/false.