Knowledgebase for Antimicrobial Resistance from machine-learned uNitigs with Integrated evidence Tiering.
KANIT is a knowledge base of genomic biomarkers associated with antimicrobial resistance in the ESKAPEE pathogens. The biomarkers are unitigs, paths of a compacted de Bruijn graph built from bacterial genomes, that gradient-boosted models select under lineage-aware cross-validation. Each biomarker is examined by independent lines of evidence, which a fixed rule combines into one evidence grade; a grade states association with resistance, not that the biomarker causes it.
Status: under development. The first release, v1.0.0, will include the knowledge base, the analysis workflow and the methods documentation.
| Path | Contents |
|---|---|
scripts/ |
the numbered analysis steps and the knowledge-base tools |
scripts/lib/ |
shared code: configuration, registries, schema, I/O |
config/ |
configuration, and the organism and antibiotic registries |
tests/ |
unit and smoke tests |
*.def, environment*.yml |
container definitions and their environments |
slurm/ |
HPC job scripts |
pip install -e ".[dev]" # or: conda env create -f environment.yml
pytest # unit and smoke testsThe code is released under the MIT License (see LICENSE).