Ask not what data standards can do for you but what you can do for data standards: a personal view of reporting standardisation in metabolomic experiments

Anyone who has written a grant, or gained public sector funding, recently in the UK, and many other countries around the globe, will be familiar with the drive to make data available through open access databases. As experiments produce ever larger datasets in both the number of variables and experimental samples examined, it becomes increasingly important that in order to maximise the recovery of information from these datasets that the wider community can view the data and associated metadata after the initial publication. There is also a growing need to be able to review these datasets alongside the manuscripts that report the data so that the reported analyses, including data processing, may be fully assessed. Finally, funders are aware that there is a lot of repetition in research, and so if datasets could be made available that were transferable between labs this should serve to reduce some preliminary experiments that may be needed as part of a larger study. One example from my own area of research is work on the ob/ob mouse—a model so well recognized of type II diabetes and obesity that many include a cohort of these animals alongside newly generated models of the metabolic syndrome as a positive control. A recognized standard dataset for this mouse may reduce animal numbers in experiments, with attendant ethical implications, as well as reduce the overall costs of experiments. For my grants I know this is relatively straightforward for my transcriptomic datasets. MIAME (Brazma et al. 2001) is reasonably well developed and my only question is whether I opt for the ArrayExpress developed by the European Bioinformatics Institute (EBI) or GEO Profiles (Gene expression Profiles) maintained by the National Center for Biotechnology Information (NCBI), USA. They share a reporting structure and my choice comes down to one of ease. For proteomics the situation is slightly trickier as while there is MIAPE (Orchard et al. 2004), with recent developments in proteomics it may be that my deposited data doesn’t quite capture the structure of the experimental design. However, the situation is far from ideal in metabolomics. I normally write something along the lines that there are currently no recognized global databases for the repository of metabolomics data, but as I was involved in the Metabolomics Standards Initiative (MSI) I should be well placed when the data standards are agreed upon and databases are produced. This is not a very satisfactory answer as some of my reviewers have been quick to point out! The data standards initiative in metabolomics, MSI, was born out of a desire to standardize the reporting of a metabolomics experiment rather than prescribe how to perform the actual experiments. It followed a number of well received publications in the field which detailed how one might go about describing a metabolomics experiment in different biological fields, and culminated in the publication of an overview document in Nature Biotechnology (Sansone et al. 2007) and a special edition of the Metabolomics journal which provided more detail of how to describe individual aspects of a metabolomics experiment, such as animal husbandry, plant cultivation, microbial culture, metabolite extraction, analysis and data processing [Metabolomics (2007) volume 3, issue 3, pp. 175–256]. However, after these initial publications progress slowed, and although some have attempted to make their data and metadata (descriptions of how the data were acquired and the experiment performed) available, the community as a whole has not yet adopted MSI. J. L. Griffin (&) Department of Biochemistry and Medical Research Council Human Nutrition Research, Cambridge, UK e-mail: Jlg40@cam.ac.uk; jlg40@mole.bio.cam.ac.uk