| name | kaiju |
| description | Fast sensitive taxonomic classification via protein-level matches. Use this skill when working with kaiju: installing, choosing parameters, running commands, interpreting outputs, or troubleshooting. Upstream: https://github.com/bioinformatics-centre/kaiju. For stage routing use tool-selection / metagenomics-workflow.
|
| license | MIT |
| category | analysis-tools |
| tags | ["taxonomy","protein","classification"] |
| upstream | https://github.com/bioinformatics-centre/kaiju |
| stage | taxonomy |
Kaiju
Upstream: Kaiju
Citation
Menzel, P., Ng, K. L. & Krogh, A. Fast and sensitive taxonomic classification for metagenomics with Kaiju. Nat. Commun. 7, 11257 (2016). https://doi.org/10.1038/ncomms11257
See also docs/references.md.
Analytical thinking
Kaiju classifies reads by six-frame translated matches against a protein DB
(LCA of hits). Strength: remote homology / understudied clades where DNA k-mers
fail. Weakness: weaker close-species separation than good DNA classifiers.
| Tool | Signal | Prefer when |
|---|
kaiju | AA / protein | Novel / divergent lineages |
kraken2 | DNA k-mers | Speed, custom DNA DBs |
metabuli | Joint DNA+AA | Want both sensitivity and specificity |
metaphlan / singlem | Markers | Clade/SGB or unknown-species marker profiles |
Database choice dominates results. Do not merge Kaiju tables with MetaPhlAn
unlabeled.
How to run
kaiju -t nodes.dmp -f kaiju_db.fmi \
-i clean_1.fq.gz -j clean_2.fq.gz -z 16 \
-o temp/kaiju/${id}.out
kaiju2table -t nodes.dmp -n names.dmp -r species \
-o temp/kaiju/${id}.tsv temp/kaiju/${id}.out
Key parameters
| Item | Notes |
|---|
| DB | RefSeq / nr / proGenomes-derived builds — pin release |
-z | Threads |
-a / greedy modes | See upstream for sensitivity vs speed |
Rank in kaiju2table | Species/genus — state in Methods |
Decision tree
Protein-aware read taxonomy?
├─ Yes, AA sensitivity primary → kaiju
├─ Joint DNA+AA → metabuli
├─ Marker composition → metaphlan / singlem
└─ Fast DNA k-mer → kraken2 (+ bracken)
Methods caveats
- Hits depend on protein DB completeness; unclassified ≠ empty community
- Pin Kaiju + NCBI taxonomy dump versions used to build the DB
Related skills
metabuli · kraken2 · diamond · metaphlan · singlem · tool-selection
