Home

bakeoff wiki

Companion documentation for Huang et al., "Critical Evaluation of Long Read Taxonomic Profiling of the Gut Microbiome" (manuscript in preparation).

This wiki is the operational manual for the bakeoff repository: how to install the environment, fetch the data, rebuild the reference databases, and regenerate every figure and table in the paper. For the scientific findings and their interpretation, see the manuscript itself.

What this study does

Six long-read-capable taxonomic classifiers and profilers - Kraken2, Centrifuge, Centrifuger, Ganon2, Sourmash, and Sylph - are benchmarked across three dataset categories (PBSIM3 + MIMIC simulated reads, the ZymoBIOMICS D6331 mock community, and the DYN clinical cohort) on both PacBio HiFi and Oxford Nanopore reads. Each method runs under (a) its native default database and (b) a shared unified RefSeq-228 build (~69k genomes spanning archaea, bacteria, fungi, viruses), holding the reference fixed to separate database effects from algorithm effects.

Where to start

Reproduction comes in three tiers. Pick the one that matches what you actually need to do:

Tier	Audience	What you need	Start here
1 - Use the shipped tables	Render any figure, sweep thresholds, write your own notebooks against the preprocessed CSVs. No pipeline runs.	Conda env only (Installation)	Reproduction → Quick
2 - Re-process from raw reads	Re-run the profilers against the shipped reference databases - validate end-to-end, test custom parameters, or extend to new samples.	Env + DBs + reads (Data Fetch)	Reproduction → Full
3 - Rebuild / extend the reference	Update the RefSeq snapshot, add new genomes, or audit the build end-to-end.	Env + raw RefSeq genomes + taxdump	Database Build

For background and reference lookups:

If you want to...	Go to
Understand the end-to-end pipeline	Pipeline Overview
Look up a dataset, tool, or metric	Datasets · Tools and Databases · Analysis Methods
Add a new tool or dataset	Extending
Hit a snag	FAQ and Troubleshooting

The 2-minute reproduction (Tier 1)

# 1. environment (one-time)
conda env create -n bakeoff -f config/bakeoff_env.yaml && conda activate bakeoff

# 2. render a figure from the shipped analysis tables - no DBs, no reads, no profiler runs
jupyter nbconvert --to notebook --execute scripts/analysis/analysis_detection.ipynb

For Tier 2 (re-process from raw reads) and Tier 3 (rebuild the reference), see Reproduction and Database Build.

Citation

Huang et al. Critical Evaluation of Long Read Taxonomic Profiling of the Gut Microbiome. (manuscript in preparation).