How segatalab/metaml works
MetAML is a command-line toolkit for predicting microbiome phenotypes and assessing microbiome–phenotype associations. A user can select samples and features from a profile, run classification or regression with model/feature selection and cross-validation, then inspect metrics and feature-importance visualizations. The repository also includes marker-to-clade conversion and plotting utilities. The README documents cross-study analysis and public metagenome profiles; the excerpts do not establish how cross-study orchestration is implemented, so that capability is represented without guessed internal links.
Main entry points: classification.py Evaluation results · dataset_selection.py Dataset selection · plot_barchart.py Feature bar chart · plot_correlation.py Abundance correlation
Components 13
Profiles and data
- Metagenome profiles
data - Dataset selection
dataset_selection.py
Prediction analysis
- Classification
classification.py - Classification variant
classification_thomas-manghi.py - Regression
regression_dev.py - Evaluation results
classification.py
Interpretation tools
- Marker conversion
tools/markers2clades.py
Result visualization
- Feature bar chart
tools/plot_barchart.py - Abundance correlation
tools/plot_correlation.py - ROC curves
tools/plot_roccurves.py - Annotated heatmap
tools/plot_annotatedheatmap.py
Other components
- Analyst
- Plots
Connections 16
- Analyst → Dataset selection: selects data
dataset_selection.py - Analyst → Classification: runs prediction
README.md - Analyst → Classification variant: runs prediction
README.md - Analyst → Regression: runs regression
README.md - Metagenome profiles → Dataset selection: supplies profiles
README.md - Dataset selection → Classification: prepares profiles
- Classification → Evaluation results: writes metrics
classification.py - Classification → ROC curves: writes estimates
- Evaluation results → Marker conversion: reads importance
tools/markers2clades.py - Evaluation results → Feature bar chart: reads results
tools/plot_barchart.py - Evaluation results → Abundance correlation: reads importance
tools/plot_correlation.py - Evaluation results → Annotated heatmap: plots results
- ROC curves → Plots: produces curves
tools/plot_roccurves.py - Feature bar chart → Plots: produces chart
tools/plot_barchart.py - Abundance correlation → Plots: produces scatter
tools/plot_correlation.py - Annotated heatmap → Plots: produces heatmap
tools/plot_annotatedheatmap.py