16S rRNA amplicon analysis workflow (QIIME2 → R)

This repository contains an R-based downstream analysis workflow for 16S rRNA amplicon sequencing data, developed during my PhD research.

The workflow is designed primarily for host-associated low-microbial biomass samples and makes extensive use of negative controls and mock communities to assess contamination, detection limits, and data plausibility.

It also makes use of environmental controls - i.e. samples from adjacent biological niches in order to compare taxonomic and phylogenetic profiles of the different sample-types.

This is a research workflow rather than a polished software package, and is shared for transparency, reuse, and reproducibility.

---

Overview of the workflow

The analysis assumes that upstream processing has already been performed using QIIME2. The R script performs the following major steps:

1. Import of QIIME2 outputs

   • Feature table
   • Taxonomy assignments (SILVA and/or Greengenes)
   • Rooted phylogenetic tree
   • Sample metadata
   • Construction of a phyloseq object

2. Initial quality control

   • Removal of non-bacterial features
   • Filtering of implausible or low-prevalence taxa
   • Basic inspection of sequencing depth and feature counts

3. Contamination assessment

   • Identification and removal of contaminant ASVs using decontam
   • Use of negative controls where available

4. Mock community analysis

   • Evaluation of mock community composition
   • Determination of minimum detection thresholds using:
     • Cell-based dilution series
     • DNA-based logarithmic mock standards

5. Final filtering

   • Removal of features below empirically determined thresholds
   • Generation of the final analysis-ready dataset
   • Analysis of alpha rarefaction and sampling depth

6. Diversity analyses

   • Alpha diversity
   • Beta diversity (ordination-based analyses)

7. Taxonomic summaries

   • Bar plots and other relative abundance summaries
   • Taxonomic composition across sample groups

8. Comparative niche analysis

   • Comparison of target samples to environmental or adjacent niches

9. Phylogenetic visualisation

   • Phylograms and tree-based representations of selected taxa

---

Intended mode of use

This workflow is designed to be run interactively in R (e.g. in RStudio), executing sections sequentially and inspecting outputs as they are generated.

Several steps (particularly initial QC, contamination assessment, and mock-based thresholding) are intentionally not fully automated, as they require dataset-specific judgement and biological plausibility checks.

While the full script can be sourced end-to-end, users are strongly encouraged to step through the analysis and review intermediate results before proceeding to downstream filtering and diversity analyses.

---

Input requirements

The workflow expects the following inputs:

• QIIME2 feature table (.qza)
• Rooted phylogenetic tree (.qza)
• Taxonomy assignments (SILVA and/or Greengenes)
• Sample metadata file (TSV format)

Details on required columns in the metadata file are described in the script comments.

---

Repository structure

analysis/
  main_analysis.R      # Main analysis script

config/
  config_example.R     # Example configuration file (paths & parameters)

data/
  README.md            # Place input data here (not tracked)

output/
  README.md            # Analysis outputs are written here

---

How to run

• Clone the repository
• Copy config/config_example.R to config/config.R
• Edit file paths and dataset-specific parameters
• Run the analysis script:

source("analysis/main_analysis.R")

---

Notes and limitations

• This workflow was developed for low-biomass 16S datasets
• Some filtering steps and mock-based thresholds are dataset-specific
• Sections of the script are clearly marked where manual intervention or adaptation may be required
• Users are encouraged to read the script comments carefully before reuse.

---

Acknowledgements

This workflow draws on methods, ideas, and code patterns from multiple sources, including but not limited to:

• Callahan et al. (DADA2)
• Davis et al. (decontam)
• F1000Research microbiome analysis guidelines
• QIIME2 documentation and tutorials

Any adaptations, errors, or interpretations are my own.