SIRIUS 4: a rapid tool for turning tandem mass spectra into metabolite structure information

Dührkop, Kai; Fleischauer, Markus; Ludwig, Marcus; Aksenov, Alexander A.; Melnik, Alexey V.; Meusel, Marvin; Dorrestein, Pieter C.; Rousu, Juho; Böcker, Sebastian

doi:10.1038/s41592-019-0344-8

Cited by 1,058 publications

(1,239 citation statements)

References 41 publications

Supporting

Mentioning

1,089

Contrasting

Unclassified

Order By: Relevance

“…It is important to obtain clean MS/MS spectrum for metabolite identification, because the presence or absence of fragments has large influence to the value of scoring function in many in-silicon methods. For example, SIRIUS 18,19 takes fragmentation trees as important features to predict substructures of metabolites, which also takes all the fragments into consideration. Significance of cleaner MS/MS spectrum.…”

Section: Resultsmentioning

confidence: 99%

Deep Learning Enable Untargeted Metabolite Extraction from High Throughput Coverage Data-Independent Acquisition

Lü

Zhang

2020

Preprint

View full text Add to dashboard Cite

The sequential window acquisition of all theoretical spectra (SWATH) technique is a specific variant of data-independent acquisition (DIA), which is supposed to increase the metabolite coverage and the reproducibility compared to data-dependent acquisition (DDA). However, SWATH technique lost the direct link between the precursor ion and the fragments. Here, we propose a deep-learning-based approach (DeepSWATH) to reconstruct the association between the MS/MS spectra and their precursors. Comparing with MS-DIAL, the proposed method can extract more accurate spectra with less noise to improve the identification accuracy of metabolites. Besides, DeepSWATH can also handle severe coelution conditions. Data dependent acquisition (DDA) selects single precursor ion for fragmentation each time, which has the direct link between the precursor ion and its fragments. In contrast, data independent acquisition (DIA) often uses a wide isolation window (10 Da -25 Da) for precursor ions selection. It allows a full coverage of observable molecules but at the expense of losing the direct link between the precursor ion and the fragments. Therefore, how to establish the link is a fundamental problem when processing DIA dataset. In proteomics, methods for this problem can be divided into two categories: peptide-centric methods and spectrum-centric methods. Peptide-centric methods usually need experimental or in silico spectral database 1-3 of all known peptides for a specific biosystem. Basing on the known spectra, OpenSWATH 4 uses reverse spectrum matching to locate the targeted peptides and precursor-fragment elution curve correlation to score the confidence of the extracted MS/MS. Specter 5 can also use curve resolution to deconvolve the multiplexed MS/MS spectra. Some peptide-centric methods, such as PECAN 6 and DIA-NN 7 , can apply the peptide sequence directly to the MS/MS deconvolution. Spectrum-centric methods, such as DIA-Umpire 8 and Group-DIA 9 , detect covarying precursor-fragment group and generate pseudospectra from DIA.

show abstract

Section: Resultsmentioning

confidence: 99%

Deep Learning Enable Untargeted Metabolite Extraction from High Throughput Coverage Data-Independent Acquisition

Lü

Zhang

2020

Preprint

View full text Add to dashboard Cite

show abstract

“…Although significant advances have been made in molecular mining workflows, chemical annotation as well as classification tools [1][2][3][4]7,8,10,[14][15][16], chemical structure annotation remains the major and most challenging bottleneck in mass spectrometry-based metabolomics as most of our biological interpretations rely on annotated structures [8,53,54]. MolNetEnhancer is a workflow that combines chemical structural information retrieved from different in silico tools, thus increasing structural information retrieved and enhancing biological interpretation.…”

Section: Discussionmentioning

confidence: 99%

“…Alternatively, other in silico tools for putative chemical structural annotation (e.g. SIRIUS+CSI:FingerID) [Dührkop et al, 2015[Dührkop et al, , 2019 can also be used. 4.…”

Section: Methodsmentioning

confidence: 99%

“…In biological samples, many metabolites share molecular substructures and form structurally related molecular families (MFs) of various chemical classes, which has inspired metabolome mining tools exploiting these biochemical relationships. Indeed, since the molecular networking approach was proposed in 2012 [10], numerous complementary molecular mining workflows as well as annotation and classification tools have been introduced including SIRIUS [3], CSI:FingerID [4], MetFusion [11], MetFamily [12], and many others [1,2,7,8,[13][14][15][16][17][18][19][20][21][22] and their combined use for natural product discovery was very recently reviewed [23]. Where tandem mass spectral molecular networking efficiently can group molecular features in molecular families [10], MS2LDA can discover substructures that aid in further annotation of subfamilies and shared modifications [14].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

MolNetEnhancer: enhanced molecular networks by integrating metabolome mining and annotation tools

Ernst

Kang

Caraballo‐Rodríguez

et al. 2019

Preprint

Self Cite

104

View full text Add to dashboard Cite

Metabolomics has started to embrace computational approaches for chemical interpretation of large data sets. Yet, metabolite annotation remains a key challenge. Recently, molecular networking and MS2LDA emerged as molecular mining tools that find molecular families and substructures in mass spectrometry fragmentation data. Moreover, in silico annotation tools obtain and rank candidate molecules for fragmentation spectra. Ideally, all structural information obtained and inferred from these computational tools could be combined to increase the resulting chemical insight one can obtain from a data set. However, integration is currently hampered as each tool has its own output format and efficient matching of data across these tools is lacking. Here, we introduce MolNetEnhancer, a workflow that combines the outputs from molecular networking, MS2LDA, in silico annotation tools (such as Network Annotation Propagation or DEREPLICATOR) and the automated chemical classification through ClassyFire to provide a more comprehensive chemical overview of metabolomics data whilst at the same time illuminating structural details for each fragmentation spectrum. We present examples from four plant and bacterial case studies and show how MolNetEnhancer enables the chemical annotation, visualization, and discovery of the subtle substructural diversity within molecular families. We conclude that MolNetEnhancer is a useful tool that greatly assists the metabolomics researcher in deciphering the metabolome through combination of multiple independent in silico pipelines.

show abstract

“…In this setup, chemicals elute through the liquid chromatographic column at different retention times (RTs) before entering the mass spectrometer and potentially undergoing fragmentation. Fragmentation produces distinct pattern of fragment peaks at different mass-to-charge ratios (m/zs) that can be used to annotate chemical structures [7,8]. The choice of fragmentation strategy, which determines how precursor ions are selected for further fragmentation in tandem mass spectrometry, is an important factor affecting the coverage and quality of MS/MS spectra available for subsequent analysis.…”

Section: Introductionmentioning

confidence: 99%

In-silico Optimisation of Mass Spectrometry Fragmentation Strategies in Metabolomics

Wandy

Davies

Hooft³

et al. 2019

Preprint

View full text Add to dashboard Cite

Liquid-Chromatography (LC) coupled to tandem mass spectrometry (MS/MS) is widely used in identifying small molecules in untargeted metabolomics. Various strategies exist to acquire MS/MS fragmentation spectra; however, the development of new acquisition strategies is hampered by the lack of simulators that let researchers prototype, compare, and optimise strategies before validations on real machines. We introduce Virtual Metabolomics Mass Spectrometer (ViMMS), a modular metabolomics LC-MS/MS simulator framework that allows for scan-level control of the MS2 acquisition process in-silico. ViMMS can generate new LC-MS/MS data based on empirical data or virtually re-run a previous LC-MS/MS analysis using pre-existing data in-silico to allow the testing of different fragmentation strategies. It allows the comparison of different fragmentation strategies on real data, with the resulting scan results extractable as mzML files. To demonstrate its utility, we show how our proposed framework can be used to take the output of a real tandem mass spectrometry analysis and examine the effect of varying parameters in Top-N Data Dependent Acquisition protocol. We also demonstrate how ViMMS can be used to compare a recently published Data-set-Dependent Acquisition strategy with a standard Top-N strategy. We expect that ViMMS will save method development time by allowing for offline evaluation of novel fragmentation strategies and optimisation of fragmentation strategy for a particular experiment.

show abstract

SIRIUS 4: a rapid tool for turning tandem mass spectra into metabolite structure information

Cited by 1,058 publications

References 41 publications

Deep Learning Enable Untargeted Metabolite Extraction from High Throughput Coverage Data-Independent Acquisition

Deep Learning Enable Untargeted Metabolite Extraction from High Throughput Coverage Data-Independent Acquisition

MolNetEnhancer: enhanced molecular networks by integrating metabolome mining and annotation tools

In-silico Optimisation of Mass Spectrometry Fragmentation Strategies in Metabolomics

Contact Info

Product

Resources

About