q2-sample-classifier: machine-learning tools for microbiome classification and regression
Problem
Microbiome researchers need to predict outcomes or differentiate samples from microbial composition, but supervised learning methods were not easily accessible or systematically benchmarked.
Why it matters
Linking microbial composition to health outcomes supports research and clinical decision-making.
What the AI did
The q2-sample-classifier plugin brings scikit-learn supervised classifiers and regressors into the QIIME 2 framework, with automated feature selection and hyperparameter tuning, and benchmarks them on microbiome data.
Results
Random forest, extra trees, and gradient boosting showed highest performance; automated feature selection and tuning enhanced most methods (no numeric metrics stated)
Impact
Makes machine-learning tools more accessible and interpretable to microbiologists, clinicians, and other researchers.
Record
| Venue | bioRxiv (Cold Spring Harbor Laboratory) |
|---|---|
| Published | 2018 |
| Work type | preprint |
| Authors | Nicholas A. Bokulich, Matthew R. Dillon, Evan K Bolyen, Benjamin D. Kaehler, Gavin A. Huttley, J. Gregory Caporaso |
| Citations | 32 (OpenAlex, at import time) |
| References | 31 |
| Model or method | scikit-learn classifiers/regressors (random forest, extra trees, gradient boosting) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1101/306167 |
| Open access | green |
| Keywords | microbiome, machine learning, supervised classification, QIIME 2, benchmarking |
Importance rating
real but narrower advance
Abstract frames the problem as accessibility of supervised learning for microbiome prediction tasks; the plugin aids research and clinical decision-making but is a tooling/benchmarking advance of narrow stated scope.