nf-pore2taxa

quick taxonomic assignment of Nanopore metabarcoding data

Introduction

nf-pore2taxa is a netxflow bioinformatics analysis pipeline used for Nanopore amplicon sequencing. Its goal is to provide quick & dirty taxonomic profiles. To do so, it supports taxonomic assignment with the sintax method, as implemented in vsearch, and produces an occurrence table (identified taxa vs. barcode ID). Read processing is minimal: a simple primer trimming and conversion to fasta. Read are then grouped based on their taxonomic assignments (sintax cutoff: 0.90). Any taxonomic reference dataset can be used (16S/18S SSU rRNA, ITS, COI, etc.), as long as it is properly formated (fasta file, with headers in sintax format).

Note

sintax format: the reference database (in fasta format) must contain taxonomic information in the header of each sequence in the form of a string starting with ;tax= and followed by a comma-separated list of up to nine taxonomic identifiers. Each taxonomic identifier must start with an indication of the rank by one of the letters d (for domain), k (kingdom), p (phylum), c (class), o (order), f (family), g (genus), s (species), or t (strain). The letter is followed by a colon (:) and the name of that rank. Commas and semicolons are not allowed in the name of the rank. Non-ascii characters should be avoided in the names.

Pipeline summary

By default, the pipeline currently performs the following:

• basecalling and demultiplexing of pod5 ('super accurate') with dorado
• trimming of reads with cutadapt
• taxonomic assignment with vsearch (sintax)
• build an occurrence table (using R and the tidyverse package)

The basecalling step can be skipped if fastq files are already available.

Usage

First, you need to prepare a config file, with the parameters for your project.

// nf-pore2taxa workflow

workDir = "/big/drive/projects/project_ID/work"

params {
    pod5_dir      = "/big/drive/runs/run_ID/pod5"
    fastq_dir     = "/big/drive/projects/project_ID/data/run_ID"
    sintax_silva  = "/safe/data/references.fasta.gz"
    results_table = "/big/drive/projects/project_ID/results/sintax.tsv"
    primer_f = "GTACACACCGCCCGTCG"
    primer_r = "CGCCTSCSCTTANTDATATGC"

    // Set to true if basecalling was already done
    // and fastq_dir already exists
    skip_basecall = false
}

Warning

Intermediate files are linked (hardlinks), so it is important for workDir and fastq_dir to be on the same filesystem.

Now, you can run the pipeline using:

nextflow \
    run main.nf \
    -config /path/to/myproject.config

Parameters can also be passed via the command-line, if need be.

Pipeline output

Two tab-separated tables with identified taxa as rows, and barcode IDs (i.e, samples) as columns. The first line is the header line (column names). An additional column with the total number of reads for each taxa is also provided (column number 2). Barcodes without any assignments are grouped at the far-right of the table.

Here is a tiny output table example with only two barcodes, and two identified taxa:

taxonomy	total	barcode03	barcode01
d:Fungi ...	1234	1234	0
d:Viridiplantae	42	42	0

barcode01 is empty (no assigned reads), so it appears last.

The second output table, marked as optimistic, has the same structure as the first table. It contains full taxonomic assignments, including assignments that are below the probability threshold (0.9).

Road-map

• eliminate variability due to sintax? not currently possible
• write unit tests using nextflow's tooling
• refactor assign_with_sintax.sh. Use nextflow to find and loop over the fastq.gz files. Operate on each file independently, publish back the results in the same directory
• add a module that checks if binaries (cutadapt, vsearch, dorado) are in PATH? already done by the different scripts
• assign_with_sintax.sh eliminate fastq to fasta conversion, cutadapt can read fastq.gz directly
• add a module that checks parameters? do we need to pass parameters that are not used? for instance, if we skip basecalling, do we need to pass the path to pod5 files?
• eliminate dependency to R and the tidyverse package, rewrite script in python?

See also

• NanoASV: a snakemake-based workflow for Nanopore amplicon sequencing (16S/18S SSU rRNA)