CFM-ID: a web server for annotation, spectrum prediction and metabolite identification from tandem mass spectra

TLDR

Competitive Fragmentation Modeling (CFM) is a probabilistic generative model that learns MS/MS fragmentation parameters from data using machine learning. CFM‑ID is a web server that automates metabolite identification by annotating spectra, predicting spectra from structures, and ranking candidate structures for a target spectrum. The server offers a simple interface with graphical displays of annotations, spectra, and structures, and is freely available at http://cfmid.wishartlab.com. Extensive testing on multiple datasets shows that CFM‑ID outperforms existing methods such as MetFrag and FingerId.

Abstract

CFM-ID is a web server supporting three tasks associated with the interpretation of tandem mass spectra (MS/MS) for the purpose of automated metabolite identification: annotation of the peaks in a spectrum for a known chemical structure; prediction of spectra for a given chemical structure and putative metabolite identification—a predicted ranking of possible candidate structures for a target spectrum. The algorithms used for these tasks are based on Competitive Fragmentation Modeling (CFM), a recently introduced probabilistic generative model for the MS/MS fragmentation process that uses machine learning techniques to learn its parameters from data. These algorithms have been extensively tested on multiple datasets and have been shown to out-perform existing methods such as MetFrag and FingerId. This web server provides a simple interface for using these algorithms and a graphical display of the resulting annotations, spectra and structures. CFM-ID is made freely available at http://cfmid.wishartlab.com.