eDNA-metabarcoding-intro

This is an introduction analyzing eDNA metabarcoding samples to evaluate diversity, assign taxonomy, and differential abundance testing. Taxonomy assignments are compared between vsearch (qiime), command line BLAST, and Tronko (a recent phylogenetic approach to taxonomy assignment).

Duplicate my qiime conda env

conda env create -f qiime2-env.yml
conda activate qiime2

Overview of pipeline

FASTQ sample QA/QC

Fastp is used to trim off the poly-G tail commonly found in amplicon nova-seq data.
Qiime imports the directory of poly-G trimmed FASTQ files into a single 'qiime file' with the 'qza' extension. Using the primer sequence, qiime's 'cutadapt' plugin removes the primer and adapters of each pair of sequences. A second 'qza' output file is created for the cutadapt trimmed data.

Denoising

Sequences can be denoised using qiime, which calls the R package 'dada2'. Denoising learns the error rate from the base call quality of the samples, and tries to fix sequencing errors when possible. Read pairs are merged into a single sequence when they sufficiently overlap and align. Denoising output is another qiime object that contains a table of the counts for each unique sequence (called ASVs, rows of table) found among the samples (columns, each sample name taken from the fastqs). The ASV sequences and the ASV ids are stored in the 'rep-seqs.qza'. The table of counts for each ASV is stored in the 'feat-table.qza' file. Both objects can be exported to a human readable format (FASTA) to visually inspect the sequences and tables. Or, qiime has a number of summary functions that can be applied to the qza files. Qiime summaries and plots can be viewed here

Taxonomy assignment

Taxonomy assignment can be performed several ways. We've found that the best taxonomy assignment strategy differs between

Qiime2 vsearch (code/qiime_tax.sh)

The reference database of rbcl for Qiime built from ____ (script for building): Qiime object taxonomy: ref-dbs/ Qiime object sequences: ref-dbs/

Tronko (code/tronko_tax.sh)

See installation instructions for tronko here. The reference database of rbcl for tronko was built from diat.barcode with [this](script for building): Tree: ref-dbs/rbcl_diat.barcode-ref-tree.txt FASTA: ref-dbs/rbcl_diat.barcode-MSA.fasta

Visualizing the taxonomy assignments

Qiime2 can generate helpful interactive barplots of the taxa abundance for sample

Optional steps:

As an additional check for the taxonomy assignments, I get the top blast hits for each ASV. If you use a specialized reference database, such as we do here, there will be many sequences with 'unassigned' taxonomy. Blasting is a way to double check that unassigned sequences are in fact off target taxa. After assigning taxonomy and blasting the sequences, I pull the results qiime and tronko taxon

jthmiller/eDNA-metabarcoding-intro