Abstract
Summary
Accurate and efficient compound annotation is a long-standing challenge for LC–MS-based data (e.g. untargeted metabolomics and exposomics). Substantial efforts have been devoted to overcoming this obstacle, whereas current tools are limited by the sources of spectral information used (in-house and public databases) and are not automated and streamlined. Therefore, we developed metID, an R package that combines information from all major databases for comprehensive and streamlined compound annotation. metID is a flexible, simple and powerful tool that can be installed on all platforms, allowing the compound annotation process to be fully automatic and reproducible. A detailed tutorial and a case study are provided in Supplementary Materials.
Availability and implementation
Supplementary information
Supplementary data are available at Bioinformatics online.
1 Introduction
Liquid chromatography coupled to mass spectrometry (LC−MS) is a comprehensive, unbiased technology to research small compounds, which has become increasingly popular in food, environment and biomedical studies (Fraga-Corral et al., 2020; Wishart 2016). This approach has been very successful however the generated data has typically not been used to its full potential in part due to an incomplete annotation of the chemicals detected. Implementing an accurate and efficient compound annotation workflow would be invaluable to assist biological hypothesis generation and data interpretation. In order to standardize data reporting from metabolomics studies, the community proposed to adopt a grading scheme for annotation confidence ranging from 1 to 4 (Sumner et al., 2007) using the following parameters: mass to charge ratio (m/z), retention time (RT) and MS2 spectral matching. Confident annotation being the Achilles’ heel of modern metabolomics experiments, recent initiatives are now focusing on generating high-quality MS2 spectral libraries that are publicly accessible. By taking advantage of these resources, some tools (Chaleckis et al., 2019) were developed to facilitate compound annotation. However, these tools are limited by: (i) the breadth of parameters used to compute an annotation score; previous tools typically only use m/z and/or MS2 spectra information but lack RT; (ii) the inability to combine spectral information from multiple sources including in-house and public databases in a streamlined fashion; (iii) the processing speed, previous tools normally do not allow the implementation on cluster servers. More detailed comparisons between the existed tools and metID are listed in Supplementary Table S1.
In this context, we developed a new R package, metID, particularly designed to (i) streamline the construction of users’ in-house databases from authentic chemical standards run in each laboratory and (ii) automated compound annotation pipeline. As annotation level 1, metID provides our spectral and RT data from our in-house database containing more than 1,000 authentic standards acquired in HILIC and RPLC modes (users need to use the same LC gradient to match RT). The users can also use the metID to construct their own in-house database for level 1 annotation. For level 2, metID now provides five public MS2 databases, and five MS1 databases for level 3 (Supplementary Table S2).
2 Features and methods
Using metID, users can easily build their in-house databases using the data from authentic standards that were acquired in their own laboratories. The public databases can also be easily organized as the database format for metID. The in-house databases in our laboratory and several public databases are provided as reference. metID can be installed on all platforms (Mac OS, Windows and Linux).
2.1 Database construction for metID
The users can use the in-house library embedded in metID, which requires the users to use the exact same LC/MS instrumental setting as metID provided. If users have in-house standards which have been acquired with RT and MS2 spectra, it is possible to build their own in-house database using the ‘construct_database()’ function (Fig. 1a). Three items are contained in the database: (i) database information, (ii) standard information and (iii) MS2 spectrum for each compound. If the users want to use the in-house library embedded in metID but without RT information, they can set the ‘rt.match.tol’ beyond your LC gradient time range.
Fig. 1.
The overview of (a) in-house database construction and (b) compound annotation in different levels
2.2 Compound annotation
RT may shift in different acquisition batches. Therefore, it is necessary to correct the RT in databases using the ‘correct_database_rt()’ function if you spike internal standards into standards and subject samples. Then metID can annotate compounds with different levels according to databases (in-house database, level 1; public MS2 database, level 2; MS1 database, level 3). For m/z, RT and MS2 spectra match scores, they are combined as one total score and scaled to 0–1 (Shen et al., 2019; Tsugawa et al., 2015) (Fig. 1b and Supplementary Material).
3 Case study
We applied metID on a published study from our lab (Contrepois et al., 2020) as a case study to demonstrate the value of metID for automatic metabolite annotation (only ∼2 h running on a Mac with 6-core 32 G memory). The authors reported 463 annotated metabolites with level 1–2 by manual inspection of m/z, RT and MS2 (Level 1 as golden standards). By using metID, we successfully annotated 942 metabolites with high confidence (level 1–2). Comparing the annotation results from metID with the original annotation in the publication, metID retrieved all the 463 annotated metabolites in the previous paper in an automatic way. Besides these overlapping annotations, metID can annotate 479 new metabolites that were not annotated in the original publication (Supplementary Fig. S5). These results indicate that metID is not only a valid approach for compound annotation with high accuracy (100%) and high speed (∼2 h) but also a powerful tool to largely increase annotation coverage in an automatic way (Supplementary Material).
4 Conclusion
metID is used for compound annotation in untargeted LC−MS-based data (metabolomics and exposomics). It allows users to build and share their in-house databases and search against all the major spectral databases. To our best knowledge, it is the first R package to allow users to build in-house databases and combine them with the public databases for compound annotation and can be applied in all computer platforms automatically. As R is very popular in the bioinformatics field, it means that as an open-source tool, metID can contribute to all community members to increase its public databases and add new methods and steps to it.
Financial Support: none declared.
Conflict of Interest: M.S. is a co-founder and member of the scientific advisory boards of the following: Personalis, SensOmics, Filtricine, Qbio, January, Mirvie and Oralome.
Supplementary Material
Contributor Information
Xiaotao Shen, Department of Genetics, Stanford University School of Medicine, Stanford, CA 94304, USA.
Si Wu, Department of Genetics, Stanford University School of Medicine, Stanford, CA 94304, USA.
Liang Liang, Department of Genetics, Stanford University School of Medicine, Stanford, CA 94304, USA.
Songjie Chen, Department of Genetics, Stanford University School of Medicine, Stanford, CA 94304, USA.
Kévin Contrepois, Department of Genetics, Stanford University School of Medicine, Stanford, CA 94304, USA.
Zheng-Jiang Zhu, Interdisciplinary Research Center on Biology and Chemistry, Shanghai Institute of Organic Chemistry, Chinese Academy of Sciences, Shanghai 200032, China.
Michael Snyder, Department of Genetics, Stanford University School of Medicine, Stanford, CA 94304, USA.
References
- Chaleckis R. et al. (2019) Challenges, progress and promises of metabolite annotation for LC–MS-based metabolomics. Curr. Opin. Biotechnol., 55, 44–50. [DOI] [PubMed] [Google Scholar]
- Contrepois K. et al. (2020) Molecular choreography of acute exercise. Cell, 181, 1112–30.e16. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Fraga-Corral M. et al. (2020) Analytical metabolomics and applications in health, environmental and food science. Crit. Rev. Anal. Chem., 10, 109. [DOI] [PubMed] [Google Scholar]
- Shen X. et al. (2019) Metabolic reaction network-based recursive metabolite annotation for untargeted metabolomics. Nat. Commun., 10, 1516. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Sumner L.W. et al. (2007) Proposed minimum reporting standards for chemical analysis. Metabolomics, 3, 211–221. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Tsugawa H. et al. (2015) MS-DIAL: data-independent ms/ms deconvolution for comprehensive metabolome analysis. Nat. Methods, 12, 523–526. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Wishart D.S. (2016) Emerging applications of metabolomics in drug discovery and precision medicine. Nat. Rev. Drug Discov., 15, 473–484. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.