Feat use mgf for sample spectra

.gitignore

@@ -30,6 +30,7 @@ examples/output
 # Sample data files
 examples/example_data/*.ipc
 examples/example_data/*.csv
+examples/example_data/*.mgf
 examples/example_data/*.parquet
 
 build/

Makefile

@@ -129,21 +129,22 @@ set-gcp-credentials:
 ## Run winnow train with sample data
 train-sample:
 	uv run winnow train \
-	dataset.spectrum_path_or_directory=examples/example_data/spectra.ipc \
+	dataset.spectrum_path_or_directory=examples/example_data/spectra.mgf \
 	dataset.predictions_path=examples/example_data/predictions.csv \
 	model_output_dir=models/new_model \
-	dataset_output_path=results/metadata.csv \
+	dataset_output_path=results/calibrated_dataset.csv \
 	calibrator.features.retention_time_feature.train_fraction=0.3 \
+	calibrator.features.retention_time_feature.min_train_points=2 \
 	calibrator.hidden_layer_sizes="[32, 16]" \
 	calibrator.early_stopping=false \
-	calibrator.max_iter=100 \
+	calibrator.max_iter=100
 
 ## Run winnow predict with sample data (uses locally trained model from models/new_model)
 predict-sample:
 	uv run winnow predict \
 	calibrator.pretrained_model_name_or_path=models/new_model \
 	fdr_control.fdr_threshold=1.0 \
-	dataset.spectrum_path_or_directory=examples/example_data/spectra.ipc \
+	dataset.spectrum_path_or_directory=examples/example_data/spectra.mgf \
 	dataset.predictions_path=examples/example_data/predictions.csv
 
 ## Clean output directories (does not delete sample data)

docs/cli.md

@@ -16,7 +16,7 @@ uv pip install winnow-fdr
 
 ## Quick Start
 
-Use the example HeLa single-shot subset in `examples/example_data/` (`spectra.ipc`, `predictions.csv`) which represents real instrument data and InstaNovo predictions.
+Use the example HeLa single-shot subset in `examples/example_data/` (`spectra.mgf`, `predictions.csv`) which represents real instrument data and InstaNovo predictions.
 
 ```bash
 # Train a calibrator on the example subset
@@ -105,7 +105,7 @@ Compute calibration features and write one enriched metadata CSV. Uses the same
 winnow compute-features
 
 # Paths and output file
-winnow compute-features dataset.spectrum_path_or_directory=data/spectra.ipc dataset.predictions_path=data/preds.csv dataset_output_path=results/features.csv
+winnow compute-features dataset.spectrum_path_or_directory=data/spectra.parquet dataset.predictions_path=data/preds.csv dataset_output_path=results/features.csv
 
 # De novo spectra (no ground truth): labelled=false; remove retention_time_feature if present
 winnow compute-features labelled=false '~calibrator.features.retention_time_feature'
@@ -170,7 +170,7 @@ Setting `label_column` while `label_source=sequence` (or omitting `label_column`
 ```bash
 # Sequence-derived labels (full match of sequence vs prediction)
 winnow diagnose-calibration diagnostics.label_source=sequence \
-  dataset.spectrum_path_or_directory=holdout/spectra.ipc \
+  dataset.spectrum_path_or_directory=holdout/spectra.mgf \
   dataset.predictions_path=holdout/preds.csv
 
 # Pre-computed labels (e.g. proteome mapping done offline)

docs/configuration.md

@@ -117,7 +117,7 @@ defaults:
 
 dataset:
   # Path to the spectrum data file or to folder containing saved internal Winnow dataset
-  spectrum_path_or_directory: data/spectra.ipc
+  spectrum_path_or_directory: data/spectra.mgf
   # Path to the beam predictions file
   # Leave as null if data source is winnow, or loading will fail
   predictions_path: data/predictions.csv
@@ -149,7 +149,7 @@ defaults:
   - data_loader: instanovo
 
 dataset:
-  spectrum_path_or_directory: data/spectra.ipc
+  spectrum_path_or_directory: data/spectra.mgf
   predictions_path: data/predictions.csv
 
 dataset_output_path: results/metadata.csv
@@ -361,7 +361,7 @@ defaults:
 
 dataset:
   # Path to the spectrum data file or to folder containing saved internal Winnow dataset
-  spectrum_path_or_directory: data/spectra.ipc
+  spectrum_path_or_directory: data/spectra.mgf
   # Path to the beam predictions file
   # Leave as null if data source is winnow, or loading will fail
   predictions_path: data/predictions.csv
@@ -438,7 +438,7 @@ defaults:
   - data_loader: instanovo
 
 dataset:
-  spectrum_path_or_directory: examples/example_data/spectra.ipc
+  spectrum_path_or_directory: examples/example_data/spectra.mgf
   predictions_path: examples/example_data/predictions.csv
 
 calibrator:

docs/examples.md

@@ -2,7 +2,7 @@
 
 ## Quick start with example data
 
-Try `winnow` in minutes using the bundled example subset under `examples/example_data/`: spectra (`spectra.ipc`) and de novo predictions (`predictions.csv`) from a real HeLa single-shot run, reduced for repository size.
+Try `winnow` in minutes using the bundled example subset under `examples/example_data/`: spectra (`spectra.mgf`) and de novo predictions (`predictions.csv`) from a real HeLa single-shot run, reduced for repository size.
 
 ```bash
 # Train a calibrator