SIRIUS: Identifying Unknown Small Molecules from Tandem Mass Spectra Without Reference Libraries

Automated identification of small molecules from LC-MS/MS data remains a major bottleneck in metabolomics and related fields. Applications of small molecule annotation include the detection of contaminants, environmental pollutants such as PFAS, leachables, extractables, biomarkers and (novel) bioactive compounds. Existing spectral reference libraries are far from comprehensive which prevents the successful annotation of many such small molecules. SIRIUS [1] is a leading software for small molecule annotation from MS/MS data. In contrast to spectral library search, it does not require reference measurements of the unknown molecules. For over a decade, its success has been based on sophisticated algorithms, including various machine learning methods, to provide state-of-the-art identification performance. To overcome the limitations of incomplete spectral libraries, SIRIUS searches MS/MS spectra in molecular structure databases (such as PubChem or HMDB) with millions of unique compounds [2,3]. Fragment annotation of MS/MS peaks and spectral library search help to additionally boost confidence in the molecular structure hits. However, even molecular structure databases do not contain every existing compound. To facilitate the annotation of truly unknown compounds, SIRIUS provides compound class prediction for over 2500 compound classes [4], de novo structure generation [5] and the ability to create custom structure databases, e.g. by generating potential transformation products of a target molecule. SIRIUS is a generalist tool for the automated untargeted analysis of large datasets. Still it provides options to easily adapt it to your specific use case. SIRIUS comes with a user-friendly GUI, CLI and API to allow easy workflow integration. [1] Dührkop et al., Nat Methods, 2019. [2] Dührkop et al., PNAS, 2015. [3] Hoffmann et al., Nat Biotechnol, 2022. [4] Dührkop et al., Nat Biotechnol, 2021. [5] Stravs et al., Nat Methods, 2022.