matchms

To match MS/MS spectra, you load query spectra and a reference library, standardize metadata, filter peaks, and compute similarity scores like CosineGreedy to rank candidate metabolites for identification.

You can import mass spectrometry spectra from MGF, MSP, mzML, mzXML, and JSON or GNPS formats to perform spectral matching and identify unknown metabolites from your data.

Modified cosine similarity accounts for precursor m/z shifts during spectral matching, allowing you to identify metabolites even when precursor ions differ between query and reference spectra, unlike standard cosine greedy scoring.

Yes, you can apply standardized filters to clean precursor m/z, retention information, adducts, formulas, and peak quality before computing spectral similarity scores for metabolite identification.

Deriving chemical identifiers like InChI, InChIKey, or SMILES is optional for structural-aware matching, but you can compute spectral similarity and rank metabolite candidates using fragmentation patterns alone.

The best way to preprocess LC-MS/MS data is to import spectra, apply metadata harmonization and peak filtering, normalize peaks, and calculate similarity scores against a reference library to rank compound candidates.