Run the nanopore metabarcoding pipeline

🚀 Start now: — first launch takes a couple of minutes to build.

⏱ Estimated time: ~60–90 minutes (including pipeline run time)  •  🟡 Practical

In this practical you’ll run nanoporemetabarcoding, a pipeline built by Eco-Flow (Eco-Flow/nanoporemetabarcoding) — from raw Nanopore reads all the way to a taxonomically-annotated ASV table.

ℹ️ Not an official nf-core pipeline. nanoporemetabarcoding was scaffolded with the nf-core template and follows its conventions (module structure, config profiles, -profile docker, etc.), which is why some of the tooling will feel familiar to Part 3 of this workshop. But it isn’t part of the official nf-core pipeline collection, isn’t listed on nf-co.re, and has no tagged release yet, as it is still in active development.

What you’ll do

• Understand what the nanopore metabarcoding pipeline is and how it works
• Inspect raw Nanopore reads
• Work out what inputs the pipeline needs
• Design a samplesheet and metadata sheet for a real experimental scenario
• Run the pipeline on its built-in test data with Docker containers
• Explore the results — quality reports, BLAST hits, taxonomy, and a community matrix
• Run the pipeline on an HPC cluster, using UCL’s Myriad as a worked example
• Learn to -resume a run and change pipeline options (let’s see…)

The experiment

We’ll use a case study to motivate the design steps: DNA barcoding of adult wasps to identify their prey.

Adult wasps were collected at two sites: oak woodland and grassland. Each wasp was individually DNA-barcoded (COI) using primers that target that region and prevents host DNA amplification (WaspExF_* forward / LuthienR_* reverse) to profile its host’s diet signature. Each site’s wasp PCR products were pooled onto one Nanopore barcode — i.e. one plate = one ONT barcode = one FASTQ file, but each well inside that plate is its own wasp, told apart later by a second, set of sort sequences used as ‘tags’.

💡 What’s an exclusion primer? A wasp carries trace host DNA from its prey in the gut/cuticle residue, but its own DNA is far more abundant in the sample, so a plain universal COI primer pair would mostly just re-sequence the wasp itself. WaspExF_* primers — the same tagged primers used for demultiplexing — are designed to be less complementary to the wasp’s own COI sequence than to other (host) DNA, so amplification of the wasp’s own template is suppressed relative to that trace host signal. It’s rarely perfect: some wasp DNA usually still gets through, which is one reason checking control wells for unexpected hits (see Step 6) matters.

Within each plate, three forward tags (F1–F3) and two reverse tags (R1–R2) are combined pairwise to label individual wells — giving each wasp (and each control) a unique forward+reverse tag combination.

We’ll design the samplesheet/metadata for this scenario ourselves in Steps 3–4, then run the pipeline for real in Step 5 using the small dummy dataset.

Step 0 — Understand nanopore metabarcoding

Metabarcoding sequences a short, taxonomically-informative marker gene (here, COI) from many samples/individuals at once, then compares each sequence against a reference database to assign it to a species (or best-available taxonomic rank).

With Nanopore sequencing, an entire multi-well PCR plate is typically pooled into one sequencing run/barcode — so a second layer of “tags” (short index sequences stuck onto the primers during PCR) is needed to work out which read came from which well.

Step 1 — Inspect the raw data

The pipeline ships with a small built-in test dataset, but we will use a custom one made for this training, following the case study above. Clone the pipeline repository first:

git clone https://github.com/Eco-Flow/nanoporemetabarcoding.git
cd nanoporemetabarcoding

The raw reads live in wasp_test_data/. Nanopore FASTQs are gzipped, so peek at them with zcat (as in Part 1):

▶️ Try it — peek at a raw FASTQ

zcat test_data/test.fastq.gz | head -8

✅ Expected output

Groups of four lines per read, same FASTQ format as Illumina — but unlike the paired, fixed-length RNA-Seq reads from Part 3, Nanopore reads are **single, variable-length long reads** with no pair: ``` @ ... ATGCGT...(a few hundred bp, length varies read to read) + !%'&&$#"...(quality string, same length as the sequence) ``` </details>

▶️ Try it — count the reads

zcat test_data/test.fastq.gz | wc -l

FASTQ uses 4 lines per read, so divide the line count by 4 to get the number of reads. This file is deliberately tiny (a handful of reads per well) so the whole pipeline runs in minutes.

Also inspect the other files you’ll need:

cat wasp_test_data/primers_f.fasta      # forward primer-tag sequences
cat wasp_test_data/primers_r.fasta      # reverse primer-tag sequences
cat wasp_test_data/metadata.csv         # which tag combination = which sample

Step 2 — Explore the pipeline’s requirements

Read the pipeline’s own usage docs: docs/usage.md and the parameter block at the top of nextflow.config.

💡 What the pipeline does for you (in containers): trims/filters reads (NanoFilt), QCs raw and filtered reads (NanoPlot), demultiplexes by primer-tag (Cutadapt, twice — forward then reverse), clusters reads per well into consensus “species” sequences (amplicon_sorter), polishes each consensus (Medaka), BLASTs each consensus against your reference database, assigns taxonomy from the best hit (taxonomizr), and builds per-run abundance/presence-absence community matrices.

Step 3 — Design a samplesheet

The samplesheet links each Nanopore barcode (one plate) to its FASTQ. It has just two columns:

▶️ Try it — design samplesheet.csv for the wasp experiment

Step 4 — Design a metadata sheet and primer-tag FASTAs

The metadata sheet is what actually resolves each demultiplexed well down to a sample name. It has three columns:

⚠️ Key rule: primer_comb is a straight string match against your FASTA headers. If the tag names don’t match exactly (typos, case, extra characters), that well’s reads won’t be assigned to a sample.

▶️ Try it — design the metadata for plate01

Plate01 has 3 forward tags (F1_WaspExF_Tab1, F2_WaspExF_Tab2, F3_WaspExF_Tab3) and 2 reverse tags (R1_LuthienR_Tab29, R2_LuthienR_Tab54), giving 6 wells: one extraction blank, one positive control, one PCR blank, and 3 adult wasps netted in the Woodland site. Write out the FASTAs and the metadata rows.

💡 Always include controls. EXT_NEG (extraction blank — no tissue), POS_CON (mock DNA of a known reference species), and BLANK (PCR no-template control) are how you catch contamination and confirm the assay worked — track them through the metadata sheet exactly like a real sample.

Step 5 — Run the pipeline

Now put the theory aside and actually run the pipeline — on its small built-in test dataset (test_data/), so it finishes in minutes regardless of whether your wasp FASTQs exist yet.

Read the parameters block in nextflow.config for the full option list.

👉 Single vs double dashes matter! A single dash (-profile, -resume) is a Nextflow core option. A double dash (--input, --metadata, --outdir) is a pipeline parameter defined by nanoporemetabarcoding.

✅ What success looks like: Nextflow prints a banner, then a live list of processes as they run (NANOFILT, NANOPLOT, CUTADAPT ×2, AMPLICON_SORTER, MEDAKA, BLAST_BLASTN, ASSIGN_TAXA, …). On the tiny test data this should complete in a few minutes, plus a bit longer the first time while Docker pulls the containers.

Troubleshooting Step 5