An Interactive Web Application for Composing Nearest-Neighbor Suspect Spectral Libraries

Untargeted metabolomics produces millions of tandem mass spectrometry (MS/MS) spectra, yet only a small fraction can be matched to reference spectra in community libraries. The nearest-neighbor suspect spectral library was introduced to narrow this gap by propagating annotations from known to unknown spectra based on spectral similarity and mass differences. This approach produces large collections of "suspect" spectra that are structurally related to known compounds and can guide hypothesis generation about unknown features in metabolomics data. We present a web application that operationalizes this concept and makes the suspect library workflow easier to run. It builds on a precomputed data cache derived from large-scale GNPS living data molecular networking. Users can interactively compose new suspect sets under different filters, optionally enrich them with additional spectrum annotations, and then visualize these suspect sets and export them as MGF files. By separating the heavy data preparation from lightweight, interactive composition and export, the web application makes the nearest-neighbor suspect library more accessible and reproducible for users who do not want to manage large-scale batch workflows themselves.