| name | kneaddata |
| description | QC wrapper focused on host (and contaminant) read removal via Bowtie2. Use this skill when working with kneaddata: installing, choosing parameters, running commands, interpreting outputs, or troubleshooting. Upstream: https://github.com/biobakery/kneaddata. For routing across tools, use tool-selection. For end-to-end pipelines, use metagenomics-workflow.
|
| license | MIT |
| category | analysis-tools |
| tags | ["host-depletion","bowtie2","qc"] |
| upstream | https://github.com/biobakery/kneaddata |
KneadData
Upstream: biobakery/kneaddata
Citation
Beghini, F. et al. Integrating taxonomic, functional, and strain-level profiling of diverse microbial communities with bioBakery 3. eLife 10, e65088 (2021). https://doi.org/10.7554/eLife.65088
See also docs/references.md.
Analytical thinking
Host reads inflate false microbial signal and waste assembly/binning compute.
KneadData’s value is reproducible host subtraction with a documented Bowtie2
index. Always record which host genome build you used.
Critical for PE workflows: enable --reorder so mate order stays synchronized;
MetaWRAP, salmon, and BASALT connectivity all assume paired IDs.
Prerequisites
| Requirement | Minimum | Recommended |
|---|
| Input | QC’d PE FASTQ | From fastp |
| Host DB | Bowtie2 index | Matching species/build |
| RAM | 16 GB | 32 GB+ |
| Threads | 4 | 8–16 |
How to run
conda activate kneaddata
kneaddata \
-i1 temp/qc/${id}_1.fastq -i2 temp/qc/${id}_2.fastq \
-o temp/hr \
--bypass-trim --bypass-trf --reorder \
--bowtie2-options '--very-sensitive --dovetail' \
-db ${db}/kneaddata/human/hg_39 \
--remove-intermediate-output -v -t 8
If fastp already trimmed, --bypass-trim avoids double-trimming.
Key parameters
| Parameter | Guidance |
|---|
-db | Absolute path to Bowtie2 index prefix directory |
--reorder | Required for PE concordance |
--bypass-trim | On if upstream fastp already cleaned |
-t | Threads per sample |
Output
Paired dehosted FASTQs (names depend on KneadData version; rename consistently).
Decision tree
Host-associated matrix?
├─ Yes → KneadData with correct host index
├─ Environmental, negligible host → document skip; still keep QC
└─ Publishing human data → prefer releasing dehosted reads
Troubleshooting
| Symptom | Action |
|---|
| PE count mismatch downstream | Re-run with --reorder; compare headers |
| Java/Trimmomatic errors | Align Java/Trimmomatic versions or bypass trim |
| Almost all reads removed | Wrong host index or extremely host-dominated library |
Related skills
fastp upstream · megahit / humann / kraken2 / basalt downstream
