feat: add merged output format and parquet directory support

[BioGeek]

Summary

Add two features needed for INFlow pipeline integration:

1. Merged output format for winnow predict

• New output_format config option in predict.yaml: "split" (default, legacy) or "merged"
• When "merged", writes a single calibrated_psms.tsv with all columns (spectrum metadata + predictions + FDR metrics) instead of splitting into two CSVs
• Legacy split output (metadata.csv + preds_and_fdr_metrics.csv) remains the default — no breaking change
• Usage: winnow predict output_format=merged

2. Parquet directory support in InstaNovoDatasetLoader

• _load_spectrum_data() now accepts a directory path containing .parquet files
• Reads and concatenates all .parquet files in the directory
• Supports InstaNovo's sharded parquet output (dataset-*-test-*.parquet)
• Raises ValueError for empty directories

Motivation

The INFlow Nextflow pipeline needs:

• A single canonical calibrated PSM table to pass between processes (merged output)
• Ability to read InstaNovo's native parquet directory output (sharded parquet files)

Test plan

• 5 new tests in tests/scripts/test_predict_output.py
• All 299 tests pass (294 existing + 5 new)
• Pre-commit hooks pass (flake8, ruff, mypy, etc.)

🤖 Generated with Claude Code

[github-actions]

Coverage Report

File	Stmts	Miss	Cover	Missing
__init__.py	0	0	100%
data_types.py	4	0	100%
calibration
__init__.py	0	0	100%
calibration_features.py	316	7	97%	247–248, 445, 734, 922, 926, 1224
calibrator.py	91	15	83%	69–70, 72, 106–109, 134–135, 137, 162–163, 167, 194–195
compat
__init__.py	0	0	100%
instanovo.py	10	6	40%	12, 14–15, 17, 24–25
datasets
__init__.py	0	0	100%
calibration_dataset.py	109	17	84%	155, 169, 171, 173, 183, 196, 249, 251–252, 258–261, 263–266
data_loaders.py	275	14	94%	23, 189, 220–221, 423, 466, 858, 862, 911, 922, 1036–1037, 1073–1074
interfaces.py	3	0	100%
psm_dataset.py	25	0	100%
fdr
__init__.py	0	0	100%
base.py	58	15	74%	81, 85–86, 91, 98–99, 105, 126, 129–130, 135, 137–138, 144, 186
database_grounded.py	28	1	96%	52
nonparametric.py	25	4	84%	62, 68–69, 72
scripts
__init__.py	0	0	100%
main.py	192	150	21%	17–20, 53, 55–56, 68, 76, 86, 88–90, 92, 94–99, 104–105, 138–140, 161–163, 166, 169, 174, 176–178, 180, 182–183, 186–187, 190, 192–193, 195, 197, 199–200, 202, 205–206, 209–210, 213–214, 217–219, 221, 238–240, 242, 244, 249, 251–253, 255–256, 258–260, 265–266, 268–270, 272, 274, 276–277, 281–284, 286–287, 289–290, 292–293, 295, 312–314, 317, 320, 325, 327–329, 331–333, 335–336, 339–340, 343, 345–346, 348, 350, 352–353, 355, 358–359, 365–366, 369–370, 373–374, 377–378, 386–387, 389, 391, 393, 395, 397–399, 402, 408, 411, 415, 454–455, 492–493, 534–535, 565–566, 593–594, 624–625
utils
__init__.py	4	0	100%
config_formatter.py	53	40	24%	29, 37–38, 40–42, 44, 55, 58–60, 62–63, 66–69, 72–74, 77–78, 80, 91, 102, 113, 127–128, 130–132, 145–147, 150, 153–154, 157–158, 160
config_path.py	76	5	93%	24–26, 117–118
peptide.py	16	0	100%
TOTAL	1285	274	78%

Tests	Skipped	Failures	Errors	Time
299	0 💤	0 ❌	0 🔥	36.882s ⏱️

[BioGeek]

Additional fixes added to this PR

3. Include winnow.utils in package distribution (3894cc6)

winnow.utils was added to main but not registered in pyproject.toml packages list, causing ModuleNotFoundError when installed via pip. Also affects main branch.

4. Gracefully handle unmappable residues in ground-truth sequences (c0c9919)

When spectrum data contains a sequence column with modification notations not in residue_remapping (e.g. C+57.021), _process_spectrum_data crashed with ConfigKeyError. Now catches the error, logs a warning, and drops the column so the pipeline continues. Correct for de novo mode.

All 299 tests pass.

[BioGeek]

Added Parquet input support for predictions (1174cd9).

Both _load_predictions_without_beams and _load_beam_preds now accept .parquet files in addition to .csv, dispatching on file extension.

Motivation: The INFlow pipeline passes predictions from InstaNovo to Winnow. With the corresponding InstaNovo change (instadeepai/InstaNovo-internal#574), predictions can be saved as Parquet for faster I/O. This change lets Winnow read them without conversion.

Also previously pushed: winnow.__version__ support via importlib.metadata (9d898b2).

No breaking changes — existing CSV workflows are unaffected.

[BioGeek]

Updated docs, tests, and config for Parquet predictions input support (3b60d1f):

• Docstrings: Updated class docstring, load(), and _load_beam_preds to mention CSV/Parquet
• Tests:
  • Fixed test_load_beam_preds_raises_for_non_csv → renamed to test_load_beam_preds_raises_for_unsupported_format, now uses .txt (since .parquet is now supported)
  • Added test_beam_columns_none_load_predictions_without_beams_parquet
• Docs: Updated datasets.md, cli.md, and instanovo.yaml config comments