One of the long-standing goals in plant biology has been to link genotypic variation to natural variation in plant development and adaptive traits. From recent studies it has become clear that a complex interacting network is underlying phenotypic diversity. A major role in this regulatory mechanism is assigned to the metabolism since plants are extremely rich and variable in metabolic content profiles. Technological advances in detecting and quantifying biochemical content as well as novel experimental approaches have accelerated data generation and increased our understanding of regulatory mechanisms in plant biology. It is now clear that modern plant sciences can benefit enormously from integrated multidisciplinary approaches.

Metabolomics approaches enable the parallel assessment of the levels of a broad range of metabolites and have been documented to have great value in both phenotyping and diagnostic analyses in plants. These tools have recently been turned to evaluation of the natural variance apparent in metabolite composition. Here, we describe exciting progress made in the identification of the genetic determinants of plant chemical composition, focussing on the application of metabolomics strategies and their integration with other high-throughput technologies. Metabolomics represents an important addition to the tools currently employed in genomics-assisted selection for crop improvement.

Motivation: Metabolic profiles derived from high resolution 1H-NMR data are complex, therefore statistical and machine learning approaches are vital for extracting useful information and biological insights. Focused modelling on targeted subsets of metabolites and samples can improve the predictive ability of models, and techniques such as genetic algorithms (GAs) have a proven utility in feature selection problems. The Consortium for Metabonomic Toxicology (COMET) obtained temporal NMR spectra of urine from rats treated with model toxins and stressors. Here, we develop a GA approach which simultaneously selects sets of samples and spectral regions from the COMET database to build robust, predictive classifiers of liver and kidney toxicity. Results: The results indicate that using simultaneous sample and variable selection improved performance by over 9% compared with either method alone. Simultaneous selection also halved computation time. Successful classifiers repeatedly selected particular variables indicating that this approach can aid defining biomarkers of toxicity. Novel visualizations of the results from multiple computations were developed to aid the interpretability of which samples and variables were frequently selected. This method provides an efficient way to determine the most discriminatory variables and samples for any post-genomic dataset. Availability: GA code available from http://www1.imperial.ac.uk/medicine/people/r.cavill/ Contact: r.cavill@imperial.ac.uk; t.ebbels@imperial.ac.uk Supplementary information: Supplementary data are available at Bioinformatics online. Associate Editor: Jonathan Wren