The Chemical Translation Service—a web-based tool to improve standardization of metabolomic reports

Abstract

Summary: Metabolomic publications and databases use different database identifiers or even trivial names which disable queries across databases or between studies. The best way to annotate metabolites is by chemical structures, encoded by the International Chemical Identifier code (InChI) or InChIKey. We have implemented a web-based Chemical Translation Service that performs batch conversions of the most common compound identifiers, including CAS, CHEBI, compound formulas, Human Metabolome Database HMDB, InChI, InChIKey, IUPAC name, KEGG, LipidMaps, PubChem CID+SID, SMILES and chemical synonym names. Batch conversion downloads of 1410 CIDs are performed in 2.5 min. Structures are automatically displayed.

Implementation: The software was implemented in Groovy and JAVA, the web frontend was implemented in GRAILS and the database used was PostgreSQL.

Availability: The source code and an online web interface are freely available. Chemical Translation Service (CTS): http://cts.fiehnlab.ucdavis.edu

Contact: ofiehn@ucdavis.edu

1 INTRODUCTION

The Metabolomics Standards Initiative (MSI) proposed the use of database identifiers for publishing reports in peer-reviewed journals or in data repositories (Sumner et al., 2007), but the MSI did not specify best practice standards which identifier to use. Consequently, metabolomic data are presented by a wide variety of identifiers, mostly using publicly available databases such as KEGG, HMDB or PubChem. In other cases, authors merely use compound names without referencing to databases. Compound names are very poor descriptors (Kind et al., 2009) as names often cannot be unambiguously mapped to authentic chemical structures, either because of missing chiral information (D, L) or because each chemical structure is associated with many synonym names, some of which may also be used for other structures. In addition, no database contains all identifiers of all other repositories. For example, KEGG LIGAND is a popular biochemical pathways database, but it is incomplete for many compounds found in human organs as given in the Human Metabolome database HMDB. Although each database lists outlinks to other databases, no single database provides comprehensive mapping options to other databases, and rarely there are batch query options offered.

Analytical chemists and biochemists may not be used to standard structure codes or lack expertise for downloading databases or installing software. We here present a publicly available tool that enables researchers to quickly convert lists of compound database identifiers, including the important InChI Keys.

Software: The Chemical Translation Service (CTS) was implemented using the programming languages Groovy (v1.7) and Java (v6.20). The open source web application framework Grails 1.2.2 and freely available plugins were used for the development of web services. For data storage, the PostgreSQL database, an open-source object-relational database management system was used. Easy access from other languages and platforms is provided via SOAP (Simple Object Access Protocol) web services.

Hardware: The used hardware was a dual quad-core Intel X5450 Xeon based server with 16 GB of RAM and an SSD (Solid State Disk) storage array with three disks in a Raid-0 configuration to store the Lucene index files. The average disk throughput was 600 MB/s. An additional SAS (Serial Attached SCSI) storage array with 16 disks in Raid-6 configuration was used to store the database content.

Source Code: The source code is hosted as a Google Code Project at http://code.google.com/p/chemical-compound-repository/

Web Front End: The database is freely available under: http://cts.fiehnlab.ucdavis.edu

2 RESULTS

The CTS consists of three major services.

1. The Discovery Service detects chemicals in provided text and returns a list of chemicals as CSV, TXT, XML or PDF.

2. The Convert Service interconverts any chemical identifier into other chemical identifiers.

3. The Batch Convert Service converts multiple identifiers of the same type into multiple identifiers.

We recommend that users, especially chemists and biologists, use CTS for standardizing metabolomic reports into MSI-compliant formats before publishing data. Single identifiers can be converted such as KEGG identifier to PubChem ID, or SMILES to KEGG ID. Importantly, batch convert services are supported (Table 1), submitting multiple chemical identifiers of the same type and yielding a table of user-selected other identifiers. We currently use the CTS service for standardizing reports from different platforms in collaborative projects.

Table 1.

Example result converting seven PubChem CIDs

NAME	PubChem CID	InChI Hash Key	KEGG	CAS	ChEBI	LIPID MAPS	HMDB	Formula	Exact Mass
2-Undecanone	8163	KYWIYKKSMDLRDC-UHFFFAOYSA-N	C01875	112-12-9	17 700	LMFA12000002		C11H22O	170.167
Apigenin	5 280 443	KZNIFHPLKGYRTM-UHFFFAOYSA-N	C01477	520-36-5	18 388	LMPK12110005	HMDB02124	C15H10O5	270.053
l-valine	6287	KZSNJWFQEVHDMF-BYPYZUCNSA-N	C00183	72-18-4	16 414		HMDB00883	C5H11NO2	117.079
Valine	1182	KZSNJWFQEVHDMF-UHFFFAOYSA-N	C16436	516-06-3		LMFA01100046		C5H11NO2	117.079
Igepal CA (630)	24 775	LBCZOTMMGHGTPH-UHFFFAOYSA-N						C18H30O3	294.219
N-carbamoyl- glutamic acid	3 679 006	LCQLHJZYVOQKHU-UHFFFAOYSA-N						C6H10N2O5	190.059
Pyruvic acid	1060	LCTONWCANYUPML-UHFFFAOYSA-N	C00022	127-17-3	32 816	LMFA01060077	HMDB00243	C3H4O3	88.016

We provide different download formats including CSV, XLS, PDF and XML, in order to empower researchers to sort, compare and arrange properties of their lists in standard office software. For example, l-valine and valine differ only in the description of stereoconfiguration. Sorting via the atom connectivity part of the InChI hash keys (labeled red in Table 1) facilitates recognition of isomers or detection of doublets. It is apparent that databases deal differently with stereo configurations, as LipidMaps does not provide valine in its regular l-configuration whereas HMDB does and KEGG provides both valine enantiomers. Other metabolites such as pyruvate, the end product of glycolysis, is given by all databases, whereas the achiral form of N-carbamoylglutamate is not represented in any other database.

The Chemical Discovery Service annotates chemicals from any given text submitted in ASCII format by text mining. A hybrid approach using word stop-lists and fuzzy regular expression matching was used for detection of chemicals and a subsequent database matching was used for the output of all possible chemical identifiers. For example, 269 chemical names were given in the supplementary material of a published non-MSI compliant report (Sreekumar et al., 2009). These names could be read into 195 structures via synonym queries, e.g. for overlap analysis with other platforms or query of related studies. Misspelled compounds (‘nonate’) did not retrieve hits.

In comparison to the Chemical Resolver Identifier beta2 (Sitzmann, 2009), the CTS software comprises more commonly used database compound identifier inputs and allows any combination of output identifiers. Importantly, CTS also enables batch queries, a feature that is not given in other resources including the recently published MetMask tool (Redestig et al., 2010). MetMask requires installation of in-house databases and provides only limited services via web based queries. CTS maintains high-query speeds because all required data were downloaded from publicly available databases (except CAS) into a new in-house DB. Due to the proprietary nature of the Chemical Abstract Services CAS, these database identifiers are incomplete as only publicly available CAS numbers were downloaded. At present, the CTS database hosts 3.8 million entries which are indexed by InChI hash keys but continues to grow in size limited by PubChem data import speed. The CTS input databases are renewed monthly for new DB updates (e.g. HMDB, LipidMaps, KEGG).

The following advantages distinguish the CTS from other services:

• Generic compound searches over millions of compounds; can be limited by source DB query language and Lucene index.

• Regular-expression-based discovery of existing chemical names in any given text.

• Batch- or single-conversion of many publicly available chemical identifiers. Further databases (MetaCyc) are being added.

• Export of results into different formats including CSV, XLS, PDF and XML.

• Access via simple and intuitive web GUIs, optimized for the most common web browsers including Mozilla Firefox, Microsoft Internet Explorer and Apple Safari.

• Access via SOAP (Simple Object Access Protocol) web services, enabling other programming languages and platforms.

• InChI hash keys as main link between identifiers, providing structure-based indexing. InChI codes can be used in chemical databases for extensive substructure comparisons.

• Inclusion of further compound information, from monoisotope exact masses to proton donor potential, aiding identifications in mass spectrometry based metabolomic profiles.