FSIWsim: Fragment Size- and Intensity- Weighted MS2 Similarity for rapid identification and analogue annotation of known natural products in Euphorbia helioscopia L. based on LC-MS/MS data of the extract mixture.

BACKGROUND

Identification of known natural products (NPs) in complex plant extracts is essential for plant metabolomics, natural product dereplication and discovery of bioactive components from medicinal herbs.

RESULTS

We propose FSIWsim, a Fragment Size- and Intensity- Weighted MS2 similarity algorithm for NP identification in complex extracts based on LC-MS/MS data. FSIWsim computes and ranks MS2 spectral cosine similarity between query and reference spectra from a curated MS2 spectral database containing 677,356 compounds. The similarity is evaluated in a combined feature space encompassing all matched fragment ions and neutral losses, and unmatched fragment ions, with weighting by both fragment size and intensity. On the CASMI 2017 dataset, FSIWsim achieved a top-1 identification accuracy of 45.96%. Validation of FSIWsim using 43 certified natural product standards achieved a top-5 accuracy of 90.70%. Application of FSIWsim to Euphorbia helioscopia L. extract (EHL) enabled identification of 101 known NPs, including 20 confirmed using authentic standards. For those that were not correctly identified, candidates structurally similar to an identified analogue-despite having a different precursor m/z-could be retrieved from a broader chemical structure database via substructure or structure similarity searches, enabling plausible annotation of the query spectrum.

SIGNIFICANCE: FSIWsim provides a robust measure of structural similarity and serves as an effective alternative tool for identifying known NPs in complex plant extracts, while also enabling the annotation of compounds absent from the spectral database but structurally similar to identified analogues.