It's only natural: Metabolomics plugin

Skip to Navigation

Ezine

  • Published: Aug 15, 2014
  • Author: David Bradley
  • Channels: Chemometrics & Informatics
thumbnail image: It's only natural: Metabolomics plugin

Natural like

In metabolomics experiments, spectral fingerprints of metabolites with no known structural identity are detected routinely. Researchers in the UK have now taken a cheminformatics approach to this problem that lets them measure how similar these unknowns are to known natural product molecules so that they can add this factor to a better ranking of the correct structures in the results list for screening and further investigation. Image credit: Steinbeck et al

In metabolomics experiments, spectral fingerprints of metabolites with no known structural identity are detected routinely. Researchers in the UK have now taken a cheminformatics approach to this problem that lets them measure how similar these unknowns are to known natural product molecules so that they can add this factor to a better ranking of the correct structures in the results list for screening and further investigation.

Kalai Vanii Jayaseelan and Christoph Steinbeck of the Cheminformatics and Metabolism team at the European Molecular Biology Laboratory - European Bioinformatics Institute (EMBL-EBI) on the Wellcome Trust Genome Campus, near Cambridge, explain that computer-assisted structure elucidation (CASE) studies in metabolomics experiments are often used to try and unravel the spectroscopic data for metabolites. But, they explain, a one-dimensional proton NMR spectrum or a straightforward mass spectrum is rarely adequate for structure elucidation of previously unknown substances. Steinbeck has previously been successful in this regard with a raft of NMR techniques including 1D and 2D approaches, but this is not fast enough for high-throughput experiments. If extra information about the natural product nature of the unknown metabolites were taken into account, however, it might be possible to rank the many unknowns more effectively so that only fast 1D NMR or MS would be needed.

Plug and play

The researchers have now tested an evolutionary algorithm that uses a component that was trained beforehand and assigns a natural product likeness to the stream of moleculare spectra it sees, small molecule metabolites emerging from metabolomics experiments. Their algorithm has now been developed as plugins for evaluating, or rather, calculating, natural product likeness within the CASE platform known as SENECA published by Steinbeck. The plugins proved themselves against 41 small molecule (up to only fifteen heavy atoms in each) candidates picked from the Journal of Natural Products in tests with carbon-13 NMR spectroscopic data rather than proton NMR.

In fact, the system has three plugins for evaluating natural product likeness - NMRShiftDBJudge, NPLikenessJudge and AntiBredtJudge - which were pitted against each other in tests on the same spectroscopic data and in combination. Success was greater if all three judges were used for the test runs rather than a single one alone or any in pairs. They were able to obtain correct structures 36 times out of the 41 test candidates this way. The character of the five that remained "unknowns" might help the team optimize the system still further.

Efficient prediction

"We have also shown that significant improvement in overall prediction frequency and average ranking can be achieved with the application of the NP-likeness filter," the team reports. "We believe that these results can lead to improvements in CASE systems for use in metabolomics data analysis pipelines. The open-source, open-data implementation allows other researchers to contribute to or modify the package, and use their own training data for fitness evaluation."

"Our future plans are to contribute to a general and reliable system for the identification of unknown metabolites on the metabolomes of the estimated 6 million species on the planet," Steinbeck told SpectroscopyNow. "The amount of 'dark matter' in those metabolomes is vast. Current estimate about the percentage of unknown metabolites for organisms with large metabolomes (plants, fungi, certain microbes) are in the range of 50 to 80 percent."

Related Links

BMC Bioinformatics 2014, 15, 234: "Building blocks for automated elucidation of metabolites: natural product-likeness for candidate ranking"

Article by David Bradley

The views represented in this article are solely those of the author and do not necessarily represent those of John Wiley and Sons, Ltd.

Social Links

Share This Links

Bookmark and Share

Microsites

Suppliers Selection
Societies Selection

Banner Ad

Click here to see
all job opportunities

Copyright Information

Interested in separation science? Visit our sister site separationsNOW.com

Copyright © 2017 John Wiley & Sons, Inc. All Rights Reserved