| name | public-databases |
| description | Route shotgun metagenomics work to the right public databases, catalogues, and benchmark datasets. Use when: (1) Choosing SRA/ENA vs MGnify vs GTDB vs proGenomes vs CAMI, (2) Distinguishing raw archives, genome warehouses, biome catalogues, tool indexes, and gold-standard benchmarks, (3) Pinning data releases for Methods, (4) Planning downloads for UHGG, GTDB R10+, KEGG, CARD, bioBakery, Kraken, CheckM2. Child skills: sra-ena, mgnify, gtdb, progenomes, cami, kegg, biobakery-databases. For software install use setup.
|
| license | MIT |
| category | utilities |
| tags | ["database","catalogue","UHGG","GTDB","MGnify","CAMI","proGenomes","SRA","ENA","KEGG"] |
Public databases
Metagenomics “data” is not one thing. Separate these layers in Methods and
in your disk layout:
| Layer | Examples | Skills |
|---|
| Raw / submitted reads | NCBI SRA, EBI ENA | sra-ena |
| Analysed studies + biome MAG/protein catalogues | MGnify, UHGG | mgnify |
| Rank-normalized taxonomy + species reps | GTDB (e.g. R10-RS226) | gtdb · gtdbtk |
| Uniformly annotated genome warehouse | proGenomes4 | progenomes |
| Gold-standard benchmarks / simulators | CAMI / CAMI II, CAMISIM | cami |
| Pathway / ortholog knowledge bases | KEGG, eggNOG, MetaCyc | kegg · eggnog-mapper · biobakery-databases |
| Tool runtime indexes | Kraken, CheckM2, ChocoPhlAn, CARD | tool skills |
| Genome language model corpora / checkpoints | OMG, OpenGenome2, gLM2 / NT / DNABERT-2 / Caduceus / Evo | genome-language-model · omg · evo2 |
| Protein language model checkpoints | ESM-2, ESM C, SaProt, ProstT5 / ProtT5, PLMSearch | protein-language-model · esm · esmc · saprot |
| Structure search indexes | AFDB, PDB, ESM Atlas (via Foldseek/Folddisco) | foldseek · folddisco · metagenomics-llm |
| Vector / embedding homology DBs | ERAST public vector DB | erast · homology-search |
| Phylogeny-compressed genome search sets | 661k / MOF collections | phylign · homology-search |
Analytical thinking
Ask before downloading:
- Claim type? Reanalysis of public runs ≠ MAG novelty vs UHGG ≠ CAMI tool
ranking ≠ KO pathway reconstruction.
- Primary resource vs tool-bundled index? GTDB release files ≠
GTDBTK_DATA_PATH; UHGG genomes ≠ a Kraken UHGG DB; KEGG website ≠ a
redistributable local dump.
- Which release? Pin accession / catalogue version / GTDB release /
CAMI dataset DOI + download date (+ checksum).
Map by purpose
| Purpose | Prefer | Skill |
|---|
| Public shotgun reads | SRA / ENA | sra-ena |
| Analysed studies + biome MAG catalogues (UHGG, …) | MGnify | mgnify |
| Bacterial/archaeal taxonomy release | GTDB | gtdb (+ gtdbtk) |
| ~2M QC'd annotated prokaryotic genomes | proGenomes4 | progenomes |
| Benchmark with gold standards | CAMI / CAMI II | cami |
| Marker taxonomy / pathway indexes | ChocoPhlAn, UniRef, MetaCyc | biobakery-databases |
| KO / KEGG pathways (licensed access) | KEGG | kegg |
| Human/gut metabolic knowledge | VMH | vmh |
| k-mer classification indexes | Kraken2 Standard / custom | kraken2 |
| Orthology (eggNOG) | eggNOG | eggnog-mapper |
| AMR reference | CARD | rgi |
| MAG completeness models | CheckM2 DB | checkm2 |
| Host depletion indexes | Bowtie2 host genomes | kneaddata |
| BASALT refinement weights | BASALT_WEIGHT | basalt / setup |
Decision tree
Need public metagenomics data/reference?
│
├─ Raw FASTQ for a study / BioProject
│ → sra-ena
├─ Analysed MGnify study or biome MAG/protein catalogue (UHGG, …)
│ → mgnify
├─ GTDB taxonomy tables / reps / classify my MAGs
│ → gtdb / gtdbtk
├─ Large consistently annotated genome set / pan-genomes
│ → progenomes
├─ Gold-standard benchmark or simulator
│ → cami
├─ MetaPhlAn / HUMAnN indexes
│ → biobakery-databases
├─ KEGG KO / pathways (check license)
│ → kegg
├─ Kraken2 / Bracken indexes
│ → kraken2
├─ eggNOG / CARD / CheckM2
│ → eggnog-mapper / rgi / checkm2
├─ Genome language model corpus / gLM2
│ → omg / genome-language-model
├─ Protein language model checkpoints (ESM / ESMC / SaProt / ProstT5)
│ → protein-language-model / esm / esmc / saprot / prostt5
└─ Unsure
→ this skill + tool-selection
Provenance checklist (pin in Methods)
- Resource name + release / catalogue version / dataset ID
- Download URL or DOI + date (+ checksum)
- Any filters (completeness, biome, habitat, ANI)
- Whether you used primary files vs a tool index built from them
Hard rules
- Never merge MetaPhlAn, Kraken, GTDB, and MGnify labels into one unlabeled table
- MGnify/UHGG membership ≠ detection in your sample without abundance evidence
- CAMI rankings ≠ production defaults for every biome
- Prefer official FTP/API; verify mirrors with checksums
- Respect KEGG and other license-restricted resources
Related skills
sra-ena · mgnify · gtdb · progenomes · cami · kegg ·
biobakery-databases · omg · genome-language-model ·
protein-language-model · metagenomics-llm · setup · tool-selection ·
kraken2 · metaphlan · humann · eggnog-mapper · rgi · gtdbtk ·
checkm2
