Standardizing the standards

One of the most daunting aspects of using genomic technologies—including microarray, proteomic, metabolomic, and other approaches—is the sheer quantity of data that they produce. With thousands of biologically relevant molecules surveyed across (increasingly) large numbers of samples, interpretation of the data requires the use of computational approaches. And while many researchers thought that storing the data could simply build on our experiences with genome sequencing, it quickly became apparent that if one was to make sense of the results from any analysis, there was a need to store much more complex ancillary data than would be necessary for genome sequence. In 1999, as microarrays were establishing themselves as a truly viable technology, the Microarray Gene Expression Data Society (MGED; http:// www.mged.org) arranged to define the critical information necessary to effectively analyze a microarray experiment and to describe a means of encoding that information. Through a series of discussions between interested parties, public presentations, and working group meetings, what emerged were the Minimal Information About a Microarray Experiment (MIAME) (Brazma et al, 2001; Ball et al, 2002, 2004) and MAGE-ML (Spellman et al, 2002), an XML-based markup language used for describing a microarray experiment. The early success of MIAME and its widespread adoption by scientific journals also exposed some of its weaknesses, including the need to develop domain-specific extensions of MIAME to capture information about the experimental design and sample characteristics necessary for interpreting data coming, for example, from toxicology experiments (MIAME-Tox; Sansone et al, 2005) and extensions to other domains such as in situ hybridizations (MISFISHIE, the Minimum Information Specification For In Situ Hybridization and Immunohistochemistry Experiments; http://scgap. systemsbiology.net/standards/misfishie). In fact, the MGED subgroup on Reporting Structure for Biological Investigations Working Groups (RSBI WGs; http://www.mged.org/ Workgroups/rsbi/rsbi.html) is looking at ways to extend MIAME to a wide range of other areas. The principles underlying MIAME, particularly the need to clearly describe an experiment and report the variables necessary for data analysis, have resonated beyond the microarray community. For example, the metabolomics/ metabonomics community/communities (I am not going to decide which is right, and by not doing so invite the scorn of all rather than one group or the other) are moving toward standardization and reporting of metabolic analyses (Lindon et al, 2005) and practitioners of proteomics have at least two XML-based standards for reporting proteomics from which to choose, HUP-ML (Hermjakob et al, 2004) and AGML (Stanislaus et al, 2004), as well as guidance through the Minimum Information About a Proteomics Experiment (MIAPE) (Orchard et al, 2004). A recent paper by Novère et al extends the reporting standards notion beyond the experimental world, to the description of quantitative models of biochemical systems and attempts to reconcile some of the various standards that have evolved. The Minimum Information Requested In the Annotation of biochemical Models (MIRIAM) standard proposed by this group is an attempt to bring together CellML (Lloyd et al, 2004) and SBML (Finney and Hucka, 2003; Hucka et al, 2003) and to gain acceptance from databases that archive models to provide access to these in a standard machinereadable format. This is an ambitious, but important, goal as systems biology hopes to produce quantitative models of cells and cellular processes. However, unless these models, which can become quite complex, are easily testable and comparable, they will ultimately be of little value. This is an important first step in helping to establish modeling and the value it will bring to developing a predictive biology, but the ultimate impact will depend on how widely the standard is adopted and how many software tools are developed to facilitate its use. The utility of XML-based standards for facilitating data analysis in the complex realms was recently highlighted in a publication by Keller et al (2005). Keller describes the trans-proteomic pipeline, a proteomics data analysis pipeline consisting of a variety of software tools, which use different open XML standards to describe the data and manage the workflow: mzXML (Pedrioli et al, 2004) for the raw mass spec data, pepXML (http://www.matrixscience.com/xmlns/schema/ pepXML_v18) for the peptides identified from the raw data, and protXML (http://sashimi.sourceforge.net/schema_ revision/protXML/protXML_v3.xsd). This pipeline serves as a converter from one format to the other, and an interpreter and integrator of the results. Whereas this may seem trivial to those of us who remember the early days of DNA sequencing, where much of what we did in analyzing data was to convert sequence formats from GenBank to FASTA to GCG to Intelligenetics and back in all iterations, what Keller’s pipeline does is much more subtle—it strings together descriptions of very different domains in the analysis, linking the spectral data