ms2function 0.1.0

Deep learning model for MS2 data annotation

MS2Function

MS2Function provides MS2 spectrum annotation and metabolite set analysis powered by MS2BioText models. Large model/data assets are hosted on Hugging Face Hub and are downloaded automatically on first use.

Install

pip install ms2function

Or for local development:

pip install -e .

Assets (Hugging Face Hub)

Default asset repo: cgxjdzz/ms2function-assets

The package looks for the following files:

• models/best_model.pth
• models/config.json
• data/hmdb_subsections_WITH_NAME.jsonl
• data/all_jsonl_embeddings.pt

Environment overrides:

• MS2FUNCTION_ASSET_DIR (use local assets directory)
• MS2FUNCTION_ASSET_REPO (override HF repo id)
• HUGGINGFACE_HUB_TOKEN or HF_TOKEN (for private repos)

Quickstart (single spectrum)

from pathlib import Path
from MS2Function import run_single

json_input = {
    "peaks": [[100.1, 200.0], [150.2, 300.0]],
    "precursor_mz": 250.3,
}

result = run_single(json_input, project_root=Path(r"d:\NTU\MS2Function"))
print(result)

Quickstart (metabolite set)

from pathlib import Path
from MS2Function import run_set

result = run_set(
    r"d:\path\to\your.csv",
    project_root=Path(r"d:\NTU\MS2Function"),
)
print(result)