How Large Is the Metabolome? A Critical Analysis of Data Exchange Practices in Chemistry

Tobias Kind, Martin Scholz, Oliver FiehnPLoS 2009 · DOI

Abstract

We conclude that present databases are not capable of comprehensively retrieving all known metabolites. Metabolome lists are yet mostly restricted to genome-reconstructed pathways. We suggest that providers of (bio)chemical databases enrich their database identifiers to PubChem IDs and InChIKeys to enable cross-database queries. In addition, peer-reviewed journal repositories need to mandate submission of structures and spectra in machine readable format to allow automated semantic annotation of articles containing chemical structures. Such changes in publication standards and database architectures will enable researchers to compile current knowledge about the metabolome of species, which may extend to derived information such as spectral libraries, organ-specific metabolites, and cross-study comparisons.

Introduction: The Data Exchange Crisis

In the genomic era, sequencing a genome allows for the reconstruction of metabolic pathways. However, these in-silico models are often incomplete, missing products from orphan genes or unannotated enzymes. To achieve a comprehensive understanding of a species, an experimentally verified list of its metabolites—the metabolome—is essential.

Currently, defining the metabolome is hindered by fragmented, non-interoperable databases and outdated publishing practices. Unlike genomics, which benefits from unified standards like GenBank, chemistry faces a "data exchange crisis." Chemical information is trapped in proprietary silos or "analog" formats like PDF-embedded images. Using rice (Oryza sativa) as a case study, researchers have highlighted how this leads to massive information loss, as scientists must rely on error-prone Optical Character Recognition (OCR) to recover data from "ground up" publications.

Methodology

To assess the current chemical data ecosystem, a "stress test" was conducted using rice (Oryza sativa) as the model organism. Eleven major databases were queried across three categories:

  1. Compound Databases: PubChem, CAS (SciFinder), Beilstein, Dictionary of Natural Products (DNP), and specialized repositories like SetupX and Dr. Duke’s database.
  2. Pathway/Ortholog Databases: KEGG, RiceCyc, and Reactome.
  3. Patent Databases: IBM Patent Search.

The evaluation focused on the number of unique compounds, data "cleanliness" (endogenous metabolites vs. external chemicals), and the availability of machine-readable identifiers for cross-referencing.

Key Insights: The "Hamburger-to-Cow" Problem

A central challenge in modern publishing is the "Hamburger-to-Cow" problem. Experimental data is born digital in the laboratory but is "ground up" into flat PDFs or bitmap images for publication (the hamburger). Subsequent researchers then attempt to use software to reconstruct the original digital structure (the cow). This process is inherently lossy and inefficient.

Furthermore, database searches often suffer from significant "noise." A search for "rice metabolites" frequently returns hits for pesticides (e.g., Malathion), synthetic drugs (e.g., Omeprazole), or fungal contaminants (e.g., Bisbynin). Without manual curation, it is nearly impossible to programmatically distinguish what a plant naturally produces from what has been applied to it externally. To combat this, the authors advocate for the universal adoption of the International Chemical Identifier (InChI) and its hashed version, the InChIKey, which acts as a digital "barcode" for molecules.

Results and Analysis

The study revealed massive discrepancies across repositories:

  • Compound Counts: The Dictionary of Natural Products (DNP) returned only 55 metabolites for rice, whereas the IBM Patent Search identified 9,780 unique structures.
  • Interoperability: Most databases failed to cross-reference standard identifiers like PubChem IDs or provide machine-readable structures (SDF or CML files), preventing automated comparisons.
  • Data Quality: Only four databases (SetupX, Dr. Duke, DNP, and KNApSAcK) provided "clean" lists of endogenous metabolites. Others were heavily contaminated with xenobiotics.
  • Resource Intensity: It is estimated that building a comprehensive, curated metabolome for rice would require the manual re-analysis of approximately 50,000 journal articles.

Future Directions

To resolve these issues, the authors advocate for a shift in publication standards and database architecture:

  • Mandatory Metadata: Journals should require the submission of machine-readable chemical structures and spectra (e.g., PubChem CIDs or InChIKeys) as a condition of publication.
  • Semantic Annotation: Transitioning to "next-generation" documents (such as Chem4Word) that embed chemical data directly, making papers automatically searchable by structure.
  • Knowledge Bases: Establishing authoritative, species-specific repositories that integrate genomic predictions with experimental evidence.

By moving away from "analog" silos toward a machine-readable, interoperable ecosystem, the scientific community can finally begin to map the true scale of the metabolome.

Summary updated

Abstract

We conclude that present databases are not capable of comprehensively retrieving all known metabolites. Metabolome lists are yet mostly restricted to genome-reconstructed pathways. We suggest that providers of (bio)chemical databases enrich their database identifiers to PubChem IDs and InChIKeys to enable cross-database queries. In addition, peer-reviewed journal repositories need to mandate submission of structures and spectra in machine readable format to allow automated semantic annotation of articles containing chemical structures. Such changes in publication standards and database architectures will enable researchers to compile current knowledge about the metabolome of species, which may extend to derived information such as spectral libraries, organ-specific metabolites, and cross-study comparisons.
BibTeX
@inproceedings{kind2009how,
  author = {Tobias Kind and Martin Scholz and Oliver Fiehn},
  title = {{How Large Is the Metabolome? A Critical Analysis of Data Exchange Practices in Chemistry}},
  booktitle = {PLoS},
  year = {2009},
  doi = {10.1371/journal.pone.0005440},
}
Figure