Quickstart
This guide walks you through running your first lipid prediction with LipiDetective.
1. Copy the Config Template
$ cp config/config_templates/config_transformer.yaml my_config.yaml
2. Edit the Configuration
Open my_config.yaml and set the file paths for your data:
model: 'transformer'
files:
predict_input: 'raw/my_spectra.mzML' # Your input mzML file
saved_model: 'lipidetective_model.pth' # Pre-trained model
output: 'my_first_run' # Output directory
workflow:
predict: True
load_model: True
See Configuration for all available options.
3. Run the Prediction
$ uv run lipidetective --config my_config.yaml
4. View the Results
Results are written to a timestamped subdirectory inside the output path you specified. The prediction CSV contains identified lipid species with confidence scores:
experiments/my_first_run/LipiDetective_Output_2026_02_08_14_30_00/
├── predictions.csv # Lipid identifications
└── ...
See Output & Metrics for a detailed explanation of all output files.
Next Steps
Configuration — Customize all settings
Data Preparation — Prepare your own training data
Workflows — Train, validate, test, and tune models