project-eDNA-yellow

Environmental DNA (eDNA) metabarcoding analysis of water samples collected at Yellow Island (San Juan Islands, WA). This repository contains the bioinformatics pipeline that takes raw amplicon sequencing reads through quality control, trimming, and taxonomic classification with QIIME 2.

Note

TODO — study description. Fill in the biological context that cannot be inferred from the code: the target marker gene (e.g. 18S rRNA / COI), PCR primers, sampling design and dates, number of biological vs. technical replicates, and the overall project goal/hypothesis. Sample filenames use Yell_July18_... while the raw-read host directory is named eDNA-yellow_island-2023 — confirm the actual collection year.

Study at a glance

Location	Yellow Island, San Juan Islands, WA
Sites	Site 1 (n = 28) and Site 5 (n = 26) — 54 samples total
Assay	eDNA amplicon metabarcoding (marker: TODO)
Sequencing	Paired-end Illumina, L001_R[12]_001.fastq.gz
Reference DB	SILVA 138 99% (silva-138-99)
Classifier	QIIME 2 2024.10, naïve-Bayes sklearn

Pipeline

Analyses run sequentially; each notebook writes to a matching directory in output/. Notebooks are authored in R Markdown and knitted to .md for viewing on GitHub.

00.00  raw reads ──► FastQC + MultiQC (raw QC)
  │
01.00  raw reads ──► fastp trim ──► FastQC + MultiQC ──► QIIME2-compliant rename
  │                                                        └► output/01.00-.../ (trimmed FastQs)
02.00  trimmed reads ──► QIIME2 import ──► quality filter ──► deblur denoise
                          ──► SILVA classify ──► taxa bar plots
                                                  └► output/02.00-qiime2/ (.qza / .qzv)

Step	Notebook	Output
Raw-read QC	code/00.00-fastqc-multiqc-raw-reads.Rmd (md)	MultiQC report
Trimming + QC	code/01.00-trimming-fastp-fastqc-multiqc.Rmd (md)	output/01.00-trimming-fastp-fastqc-multiqc/
Taxonomic classification	code/02.00-qiime2.Rmd (md)	output/02.00-qiime2/

Viewing results

QIIME 2 .qzv visualizations in output/02.00-qiime2/ cannot be rendered on GitHub. Download a .qzv file and open it at view.qiime2.org (drag-and-drop; nothing is uploaded to a server). Key files: taxa-bar-plots.qzv, table.qzv, taxonomy.qzv, deblur-stats.qzv.

Reproducing the analysis

Software

• R + RStudio (notebooks use knitr, dplyr, reticulate)
• fastp, FastQC, MultiQC
• QIIME 2 2024.10 in a qiime2-amplicon-2024.10 conda/mamba environment

Steps

1. Clone this repo.
2. Update the path variables at the top of each notebook — e.g. repo_dir, qiime2_conda_path, and qiime2_conda_env_name in 02.00-qiime2.Rmd — to match your machine.
3. Provide the reference data in data/: the SILVA reference sequences (silva-138-99-seqs.qza, tracked) and the naïve-Bayes classifier silva-138-99-nb-classifier.qza (not tracked — see data/README.md).
4. Knit the notebooks in order (00 → 01 → 02).

Note

TODO — reproducibility. Add a conda environment export (qiime2-amplicon-2024.10.yml) and an R sessionInfo() capture so the exact software versions are pinned.

Data provenance

• Raw reads are hosted on the Roberts Lab server, not in git: https://owl.fish.washington.edu/nightingales/eDNA-yellow_island-2023/ (00.00 downloads them).
• Sample metadata: data/raw-fastqs/metadata.tsv.
• See data/README.md for the full data inventory.

Repository layout

├── code/     Analysis notebooks (.Rmd + knitted .md) and references.bib
├── data/     Reference sequences, sample metadata, raw-read QC reports
└── output/   Per-notebook outputs (directory name matches the notebook)

Naming and organization conventions are documented in code/README.md and output/README.md.

Status & roadmap

The pipeline runs end-to-end and all three notebooks have been executed. Open items:

• Fill study-description and reproducibility TODOs above.
• Summarize headline results — see output/02.00-qiime2/README.md.
• Investigate low deblur retention (~7.8% of reads reach the feature table).
• Add downstream ecology: alpha/beta diversity and Site 1 vs. Site 5 comparison, with exported figures/tables (not just .qzv).
• Confirm paired-end handling in 02.00 — reads are imported as paired-end but deblur denoise-other operates on forward reads; verify R2 use is as intended.
• Untrack ~1.2 GB of intermediate data (trimmed FastQs + multiqc_data.json) — see data/README.md. (History rewrite to shrink .git still optional.)

Attribution

Part of the Roberts Lab, University of Washington. Pipeline authored by Sam White.