Tracking taxonomy updates

When to use

Use this Skill when the user asks for any of:

• “latest / most recent” taxonomy proposals, ratifications, or consensus statements
• differences between NCBI Taxonomy and GTDB
• ICTV virus taxonomy standards (realms → species; MSL updates; binomial species naming)
• how to use NCBI taxids, lineages, and TaxonKit
• sequence→taxonomy assignment using QuickClade, GTDB-Tk, EukCC, vConTACT3

Navigation

Read only what you need (progressive disclosure):

• Authoritative sources to check (by domain) → reference/sources.md
• Ranks + identifiers (NCBI taxids, “no rank”, virus realms) → reference/ranks-and-identifiers.md
• Tool cheat sheets (QuickClade, GTDB-Tk, EukCC, vConTACT3, TaxonKit) → reference/tools.md
• Environment setup (Pixi; containers) → env/README.md
• Reporting templates → reference/report-template.md
• QA checklist → reference/qa-checklist.md
• Last-verified examples (time-stamped; may be outdated) → reference/last-verified-snapshots.md

Non-negotiables

• For every “most recent” claim: include date + release/proposal identifier + authority (NCBI/GTDB/ICTV/etc.).
• Never rely on names as join keys. Prefer stable IDs (NCBI taxid; GTDB genome identifiers; ICTV MSL IDs if present).
• Always capture provenance (tool version + database release + run date + command/settings).
• Preserve no-rank clades (don’t coerce into ranks).
• When sources disagree, report both and flag conflicts explicitly.

Workflow A: Taxonomy update scan (proposals + consensus + releases)

Copy/paste this checklist and check items off in your response:

Taxonomy Update Scan
- [ ] A1. Define scope (domains + timeframe + output format)
- [ ] A2. Pull authoritative updates (NCBI/GTDB/ICTV/euk frameworks)
- [ ] A3. Extract versioned changes (what changed + date + identifier)
- [ ] A4. Identify pipeline impacts (ranks, names, IDs, joins)
- [ ] A5. Deliver a versioned summary table + action items

A1. Define scope

Default scope if user is vague:

• Domains: Bacteria, Archaea, Viruses, Eukaryota
• Timeframe: last 12 months
• Output: 1-page summary + “what breaks/changes” for pipelines

A2. Pull authoritative updates

Use reference/sources.md. Prefer:

• committee ratifications (ICTV)
• database release notes / announcements (NCBI Taxonomy, GTDB)
• consensus framework papers (eukaryotes)

A3. Extract “what changed” (versioned)

For each domain, capture:

• Authority/source (NCBI Taxonomy / GTDB / ICTV / UniEuk / etc.)
• Release / proposal / ratification identifier
• Release/ratification date (UTC date if available)
• Summary of changes (new ranks, renamed clades, reorganized lineages, binomialization)
• Known impacts (taxid merges/deletions, renamed high-level nodes, schema changes)

A4. Pipeline impacts to always flag

• Rank schema changes (e.g., domain vs superkingdom; viral realm usage).
• Name changes that don’t change the underlying taxon (synonyms/aliases).
• ID changes: taxid merges/deletions (breaks downstream joins).
• Virus species naming shifts (binomialization) that break string matching.

A5. Output format

Produce:

1. Short narrative (what changed + why it matters)
2. One table: domain × source × version/date × key changes × action items
3. Compatibility guidance (if requested): how to store both old/new labels and IDs

Workflow B: Taxonomy assignment (sequence → taxonomy)

Copy/paste this checklist and check items off:

Taxonomy Assignment
- [ ] B1. Identify target domain (bac/arch/euk/virus/mixed)
- [ ] B2. Choose tool(s): QuickClade / GTDB-Tk / EukCC / vConTACT3
- [ ] B3. Run tools and capture raw outputs + tool/db versions
- [ ] B4. Normalize & enrich with NCBI taxids (TaxonKit) where applicable
- [ ] B5. Validate (cross-check sources, flag conflicts, record uncertainty)
- [ ] B6. Deliver a standardized results table + provenance

B1. Identify target domain

Use evidence from:

• marker genes, known sample context, assembly stats
• a fast first-pass classifier (QuickClade), then confirm with domain-appropriate tools

B2. Tool selection (rules of thumb)

• Bacteria/Archaea genomes (MAGs/SAGs/isolate genomes): GTDB-Tk for genome-based taxonomy
• Eukaryotic MAGs/genomes: EukCC for QC + taxonomic context
• Viruses: vConTACT3 to cluster and support taxonomic inference; interpret with ICTV standards
• Fast screening for bins/contigs: QuickClade (k-mer based)

See reference/tools.md.

B3. Environment guidance (preferred)

• QuickClade: run in the official BBTools container bryce911/bbtools:39.65 (Docker or Apptainer).
• GTDB-Tk / EukCC / vConTACT3 / TaxonKit: manage dependencies with Pixi using env/pixi.toml.

See env/README.md.

B4. Normalize with NCBI taxids (when applicable)

Include:

• ncbi_taxid
• scientific_name
• rank
• lineage_names (semicolon-delimited)
• lineage_taxids (semicolon-delimited)

Use TaxonKit; treat merged/deleted taxid warnings as data-quality events.

See reference/ranks-and-identifiers.md.

B5. Validate and flag uncertainty

Flag needs-review when:

• GTDB and NCBI disagree at high ranks
• k-mer screening disagrees with marker-gene phylogeny
• viral contigs cluster ambiguously or lack hallmark genes
• taxids are merged/deleted or names are synonyms causing joins to fail

B6. Standard output schema

Use the template in reference/report-template.md.

References and templates

• Sources: reference/sources.md
• Tools: reference/tools.md
• IDs/ranks: reference/ranks-and-identifiers.md
• Environment: env/README.md
• QA: reference/qa-checklist.md